Speed up empty-database full indexing

[Widthdom]

Summary

• Reduce the cost of the first full index on an empty database across extraction, worker handoff, reference facts, and graph candidate construction.
• Materialize C# instantiation-family facts once and reuse indexed TEMP facts while preserving partial, overload, optional/default/params, orphan, value-type, and ambiguity semantics.
• Keep the change set documented in English and Japanese with 11 focused commits and matching tests.

Validation

• dotnet build CodeIndex.sln -c Release -m:1 --no-restore -p:UseSharedCompilation=false (0 warnings, 0 errors)
• Release solution tests: net8 12,108 passed / 0 failed / 9 conditional skips; net9 11,623 passed / 0 failed / 431 conditional skips
• DatabaseTests: 351 passed on net8 and net9
• dotnet format CodeIndex.sln --no-restore --verify-no-changes
• Changelog validation and git diff --check passed
• Independent formal adversarial review: No blocking/actionable issues found.

Benchmark

Same 1,470-file corpus, fresh external database per run, Release binaries, alternating baseline/current order (B,C,C,B,C,B):

• origin/main median: 53,349 ms
• this branch median: 46,762 ms
• improvement: 6,587 ms (12.35%)
• all runs matched 12,661 chunks, 56,662 symbols, and 528,100 references; index and reference graph were complete/current

Checklist

• The change follows the existing code style.
• The change includes tests when behavior changes.
• The change includes a changelog fragment when user-visible behavior changes.
• Documentation has been updated when needed.