fix(engine): Base64-encode Python worker startup config to survive Windows argv

[yangzhang75]

What changes were proposed in this PR?

Follow-up fix for #5597, which started passing the Python worker startup config as a
single JSON-string command-line argument. On Windows the JVM assembles argv into one
command line and the inner double quotes are stripped before Python receives them, so
json.loads fails with JSONDecodeError: Expecting property name enclosed in double quotes.
Linux/macOS are unaffected (the JVM passes argv directly, quotes survive).

This Base64-encodes the JSON on the JVM side (encodeStartupConfig) and Base64-decodes it
in texera_run_python_worker.py before parsing. Base64 uses only [A-Za-z0-9+/=], so the
argument carries no quotes or spaces and survives argv quoting on every platform. The 19-key
contract and all validation are unchanged.

Any related issues, documentation, discussions?

Closes #5916

How was this PR tested?

• Scala: PythonWorkflowWorkerStartupConfigSpec updated to decode Base64; added a regression
  test asserting the encoded output contains no quotes/whitespace. 5/5 pass; scalafmt clean.
• Python: test_run_python_worker.py updated to feed Base64 input; added a round-trip test.
  ruff check + format clean.
• Verified the Base64 round-trip is byte-compatible between Java Base64.getEncoder and
  Python base64.b64decode.

Was this PR authored or co-authored using generative AI tooling?

Generated-by: Claude Code (Claude Opus 4.8)

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• No candidates found from git blame history.

[xuang7]

It seems this branch has diverged from the current main branch. Could you clean up the branch.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 54.69%. Comparing base (c3161f7) to head (28c8df1).
⚠️ Report is 15 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #5917      +/-   ##
============================================
+ Coverage     54.50%   54.69%   +0.18%     
- Complexity     2915     2952      +37     
============================================
  Files          1108     1110       +2     
  Lines         42807    42855      +48     
  Branches       4604     4608       +4     
============================================
+ Hits          23332    23438     +106     
+ Misses        18119    18050      -69     
- Partials       1356     1367      +11     

Flag	Coverage Δ		*Carryforward flag
access-control-service	70.44% <ø> (ø)		Carriedforward from c3161f7
agent-service	34.36% <ø> (ø)		Carriedforward from c3161f7
amber	56.99% <100.00%> (+0.47%)	⬆️
computing-unit-managing-service	1.65% <ø> (ø)		Carriedforward from c3161f7
config-service	57.35% <ø> (ø)		Carriedforward from c3161f7
file-service	58.59% <ø> (ø)		Carriedforward from c3161f7
frontend	48.27% <ø> (ø)		Carriedforward from c3161f7
pyamber	90.20% <100.00%> (+<0.01%)	⬆️
python	90.76% <ø> (ø)		Carriedforward from c3161f7
workflow-compiling-service	58.69% <ø> (ø)		Carriedforward from c3161f7

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

⚠️ Benchmark changes need a look

🟢 0 better · 🔴 5 worse · ⚪ 10 noise (<±5%) · 0 without baseline

Compared against main 1c580e5 benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

	config	throughput	MB/s	latency	max Δ latest / 7d
🔴	bs=10 sw=10 sl=64	411	0.251	23,767/31,766/31,766 us	🔴 +22.9% / 🔴 +113.4%
⚪	bs=100 sw=10 sl=64	817	0.499	120,635/156,208/156,208 us	⚪ within ±5% / 🔴 +47.0%
⚪	bs=1000 sw=10 sl=64	912	0.557	1,100,507/1,127,471/1,127,471 us	⚪ within ±5% / 🔴 +13.4%

Baseline details

Latest main 1c580e5 from same runner

config	metric	PR	latest main	7d avg	Δ latest	Δ 7d
bs=10 sw=10 sl=64	throughput	411 tuples/sec	459 tuples/sec	781.13 tuples/sec	-10.5%	-47.4%
bs=10 sw=10 sl=64	MB/s	0.251 MB/s	0.28 MB/s	0.477 MB/s	-10.4%	-47.4%
bs=10 sw=10 sl=64	p50	23,767 us	22,514 us	12,542 us	+5.6%	+89.5%
bs=10 sw=10 sl=64	p95	31,766 us	25,853 us	14,886 us	+22.9%	+113.4%
bs=10 sw=10 sl=64	p99	31,766 us	25,853 us	17,580 us	+22.9%	+80.7%
bs=100 sw=10 sl=64	throughput	817 tuples/sec	812 tuples/sec	999.37 tuples/sec	+0.6%	-18.2%
bs=100 sw=10 sl=64	MB/s	0.499 MB/s	0.496 MB/s	0.61 MB/s	+0.6%	-18.2%
bs=100 sw=10 sl=64	p50	120,635 us	120,386 us	99,687 us	+0.2%	+21.0%
bs=100 sw=10 sl=64	p95	156,208 us	149,966 us	106,271 us	+4.2%	+47.0%
bs=100 sw=10 sl=64	p99	156,208 us	149,966 us	115,445 us	+4.2%	+35.3%
bs=1000 sw=10 sl=64	throughput	912 tuples/sec	924 tuples/sec	1,036 tuples/sec	-1.3%	-12.0%
bs=1000 sw=10 sl=64	MB/s	0.557 MB/s	0.564 MB/s	0.632 MB/s	-1.2%	-11.9%
bs=1000 sw=10 sl=64	p50	1,100,507 us	1,073,848 us	970,675 us	+2.5%	+13.4%
bs=1000 sw=10 sl=64	p95	1,127,471 us	1,137,266 us	1,011,928 us	-0.9%	+11.4%
bs=1000 sw=10 sl=64	p99	1,127,471 us	1,137,266 us	1,045,045 us	-0.9%	+7.9%

Raw CSV

config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,486.52,200,128000,411,0.251,23766.55,31766.35,31766.35
1,100,10,64,20,2448.25,2000,1280000,817,0.499,120635.30,156207.52,156207.52
2,1000,10,64,20,21935.23,20000,12800000,912,0.557,1100507.11,1127470.78,1127470.78

[Yicong-Huang]

@yangzhang75 please rebase your commit history on this branch. right now it contains 4461 commits.

I marked the PR into a draft for now. please reopen it when ready.

[carloea2]

Sure, feel free to ping me when is ready for a review pass. thanks.

[Yicong-Huang]

By default, PR is ready for review. If not, author should mark it a draft

[carloea2]

Ok I am doing testing now

[carloea2]

It worked, thanks.

[Yicong-Huang]

ok, approved for the change. however base64 causes the payload not readable by human. could potentially make future debugging harder. low priority now.

Thanks!

[Yicong-Huang]

as this is a fix for a refactor done in #5597, we don't need to backport it to v1.2

cc @xuang7 for context.