Fix FTS search degradation diagnostics

[Widthdom]

Summary

• Add zero-result JSON/MCP diagnostics when every literal FTS token exceeds SQLite FTS5 unicode61's 1000-character token cap.
• Report indexed files containing overlong unbroken FTS tokens as fts_token_too_long validation issues.
• Normalize literal FTS search input to Unicode NFC and keep Latin-diacritic explicit-prefix search covered by regression tests.

Note: #1458's stop-word premise does not reproduce with this repository's fts5(content, content='chunks', content_rowid='id') schema because default SQLite FTS5 has no stop-word filter. The PR keeps diagnostics to actual token filtering/degradation cases.

Validation

• dotnet build CodeIndex.sln -c Release -p:UseSharedCompilation=false
• dotnet test CodeIndex.sln -c Release -p:UseSharedCompilation=false
• dotnet test tests/CodeIndex.Tests/CodeIndex.Tests.csproj -c Release -p:UseSharedCompilation=false --filter "FullyQualifiedName~DbReaderTests|FullyQualifiedName~FileIndexerTests|FullyQualifiedName~McpServerTests|FullyQualifiedName~QueryCommandRunnerTests"
• dotnet run --project tools/CodeIndex.Changelog -- check
• dotnet ./src/CodeIndex/bin/Debug/net8.0/cdidx.dll status --check --json

Documentation / Changelog

• Updated USER_GUIDE.md.
• Added changelog.d/unreleased/1458.fixed.md.

Follow-up Candidates

• None.

Fixes #1458
Fixes #1606
Fixes #1776