Fix PostgreSQL index names and bootstrap races

[eldonm]

Summary

Generated PostgreSQL compound index names could exceed PostgreSQL's 63-byte identifier limit, causing important typed unique indexes to be skipped. Concurrent first use could also race while creating the same collection table and fail with a duplicate type error.

This change retains existing short index names, uses a stable digest for longer names while preserving the suffix, and serializes first-use schema creation per collection. It updates the PostgreSQL contract and changelog.

Validation

• Full suite with coverage and a disposable PostgreSQL 16 service
• Targeted PostgreSQL index creation/idempotence and concurrent bootstrap tests
• check yaml...............................................................Passed
  check json...............................................................Passed
  trim trailing whitespace.................................................Passed
  black....................................................................Passed
  isort....................................................................Passed
  flake8...................................................................Passed
  mypy.....................................................................Passed
  Detect Secrets...........................................................Passed

One initial full-suite run reproduced the concurrent bootstrap race; it passed after the per-collection lock was added.

[github-actions]

Benchmark comparison

Threshold: ±25% (informational, does not block merge)

benchmark	baseline (s)	current (s)	delta	status
tests/benchmarks/test_deferred_save_benchmarks.py::test_bench_deferred_save_batched_100	0.015681	0.021031	+34.1%	REGRESSION (+34.1%)
tests/benchmarks/test_deferred_save_benchmarks.py::test_bench_immediate_save_100	0.016093	0.019963	+24.0%	OK
tests/benchmarks/test_jsondb_benchmarks.py::test_bench_jsondb_batched_saves_500	0.639179	0.389805	-39.0%	IMPROVED (-39.0%)
tests/benchmarks/test_jsondb_benchmarks.py::test_bench_jsondb_count_empty_query	1.045313	0.834264	-20.2%	OK
tests/benchmarks/test_jsondb_benchmarks.py::test_bench_jsondb_count_filtered	0.930707	0.983439	+5.7%	OK
tests/benchmarks/test_jsondb_benchmarks.py::test_bench_jsondb_find_filtered	1.077336	0.743033	-31.0%	IMPROVED (-31.0%)
tests/benchmarks/test_jsondb_benchmarks.py::test_bench_jsondb_save_throughput	0.002232	0.001444	-35.3%	IMPROVED (-35.3%)
tests/benchmarks/test_sqlite_benchmarks.py::test_bench_sqlite_count_empty	0.200982	0.224410	+11.7%	OK
tests/benchmarks/test_sqlite_benchmarks.py::test_bench_sqlite_count_fallback_via_regex	0.280437	0.347547	+23.9%	OK
tests/benchmarks/test_sqlite_benchmarks.py::test_bench_sqlite_count_pushdown	0.197793	0.277714	+40.4%	REGRESSION (+40.4%)
tests/benchmarks/test_sqlite_benchmarks.py::test_bench_sqlite_find_fallback_via_regex	0.275584	0.332843	+20.8%	OK
tests/benchmarks/test_sqlite_benchmarks.py::test_bench_sqlite_find_pushdown	0.173991	0.217105	+24.8%	OK
tests/benchmarks/test_sqlite_benchmarks.py::test_bench_sqlite_sort_limit_pushdown	0.221253	0.283424	+28.1%	REGRESSION (+28.1%)