docs: add Java/Python parity map and guard legacy branding

[ambiorix2099]

Summary

Adds the Java-side counterpart to the Python SDK's documentation parity map (conductor-oss/python-sdk#441) and closes the remaining Conductor terminology gaps in this repo.

Scoping note for reviewers. PR #441's title is "align Python SDK documentation with the Java SDK's documentation structure" — that architecture originated here, and Python was catching up. Of the 43 doc files #441 added to Python, 41 already existed in this repo, and this repo already had zero AGENTSPAN_* env vars. So this is deliberately not a port of those 519 files; it is the small set of things Java genuinely lacked. The parity map now records that direction in writing so the next person doesn't attempt the inverted port.

Changes

• docs/documentation-parity.md (new) — records the shared doc sets both SDKs must keep identical, the Java-only pages (FileClient, Spring), and the Python-only pages. Critically it separates real gaps (workflow-message-queue.md, the claude-agent-sdk bridge) from superseded legacy pages (the eleven uppercase docs/*.md and four old agent pages). Without that split a reader sees 15 "missing" files and manufactures work.
• docs/README.md — links the new page under Documentation conventions so it isn't orphaned.
• ci.yml: new "Reject legacy product branding" step — scans the whole tree rather than just docs/, since the legacy name had also reached shell and Java comments. The pattern is written as a character class so the workflow file doesn't match itself.
• ci.yml: lychee --include-fragments — #anchor targets are now validated, not just file existence. Previously a link to a heading that no longer existed passed CI silently.
• Terminology cleanup — clears the legacy name from package-e2e-bundle.sh provenance notes (4 refs) and one McpTool comment.

Deliberate exclusion

The branding guard excludes <legacy>.embedded, which remains in two package-e2e-bundle.sh lines. That property is a server-side boot flag owned by orkes-conductor, not an SDK setting — no Java code here sets it. Renaming it in our docs would point users at a flag that may not exist upstream. If the server has renamed it, that's a one-line follow-up; it shouldn't be guessed at from this repo.

Verification

Each CI doc guard was run locally with its exact command before commit:

Check	Result
Retired-reference guard	passes
New branding guard	passes (only the 2 excluded flag lines remain)
Links + anchors	50 docs, 260 local links, 0 missing files, 0 missing anchors
bash -n package-e2e-bundle.sh	OK
ci.yml YAML parse	OK, new step in correct position
:conductor-client-ai:compileJava	exit 0, 0 errors

One caveat: lychee wasn't available locally, so anchors were validated with a script mimicking GitHub's slug rules rather than lychee itself. 260/260 clean is a strong signal, but if lychee's slug handling differs on an edge case, --include-fragments may surface a genuine broken anchor in CI. This PR's own CI run is the real test of that.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ	Complexity Δ
...a/org/conductoross/conductor/ai/tools/McpTool.java	54.34% <ø> (ø)	1.00 <0.00> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.