fix: Python docstring detection and snapshot JSON injection (#179)

[justrach]

Summary

From the 8 bugs in #179, 4 were already fixed in prior commits. This PR fixes the remaining 2 actionable ones:

• Python docstring detection — replaced naive std.mem.count triple-quote check with position-aware std.mem.indexOf logic. Now correctly handles: inline docstrings ("""text"""), opening docstrings with text after ("""starts here), and multi-line docstrings containing def/class lines that shouldn't be indexed as symbols
• Snapshot JSON injection — replaced raw {s} path interpolation in snapshot.zig with writeJsonEscaped(). Files with ", \, or control characters in paths no longer corrupt the snapshot cache

Already fixed (no changes needed):

• C/C++ block comment exclusion — .c/.cpp already in the language check
• u16 file count truncation — already using u32
• ANSI escape strip — already parsing full CSI sequences
• Telemetry data race — call_count already std.atomic.Value(u32) with fetchAdd

Not addressed (larger scope):

• Global mutable active_pair_freq — needs architectural change
• HTTP unbounded threads — needs thread pool, separate issue

Test plan

• All existing tests pass
• New test: inline docstring """text""" does not leak fake symbols
• New test: multi-line docstring containing def does not produce phantom symbols

🤖 Generated with Claude Code

[github-actions]

Benchmark Regression Report

Threshold: 10.00%

Tool	Base (ns)	Head (ns)	Delta	Status
codedb_bundle	670801	674821	+0.60%	OK
codedb_changes	109481	116178	+6.12%	OK
codedb_deps	30771	30606	-0.54%	OK
codedb_edit	22865	23729	+3.78%	OK
codedb_find	142034	151614	+6.74%	OK
codedb_hot	151125	151631	+0.33%	OK
codedb_outline	476000	477298	+0.27%	OK
codedb_read	151274	148726	-1.68%	OK
codedb_search	290437	292983	+0.88%	OK
codedb_snapshot	4472281	4498362	+0.58%	OK
codedb_status	254734	262690	+3.12%	OK
codedb_symbol	70201	65795	-6.28%	OK
codedb_tree	91080	87544	-3.88%	OK
codedb_word	92496	93039	+0.59%	OK