Speed up snapshot JSON generation

[justrach]

Summary

• avoid sorting outline paths twice by reusing one sorted path list for tree + outlines
• serialize the maintained Explorer.symbol_index directly instead of rebuilding a temporary symbol map for every snapshot call
• write the tree directly as escaped JSON instead of allocating a standalone tree string first
• speed up JSON string escaping by appending unescaped spans in chunks
• strengthen the snapshot JSON test to assert tree and symbol-index content

Local benchmark

Compared this branch against origin/release/0.2.579 with scripts/compare-bench.py:

• codedb_snapshot: 938750 ns -> 575900 ns (-38.65%, -362850 ns)

The benchmark corpus includes src/snapshot_json.zig, so the snapshot payload grew slightly from the code change itself, but latency still dropped substantially.

Validation

• zig build test
• zig build
• python3 scripts/run-bench-json.py /tmp/codedb-head-bench.json
• base comparison via /tmp/codedb-release-base at origin/release/0.2.579

[github-actions]

Benchmark Regression Report

Thresholds: 10.00% and 50,000 ns absolute delta

NOISE means the percentage threshold was exceeded, but the absolute delta was too small to fail CI.

Tool	Base (ns)	Head (ns)	Delta	Abs Delta (ns)	Status
codedb_bundle	479609	497116	+3.65%	+17507	OK
codedb_changes	53038	56938	+7.35%	+3900	OK
codedb_deps	9422	9470	+0.51%	+48	OK
codedb_edit	6383	6307	-1.19%	-76	OK
codedb_find	61150	69012	+12.86%	+7862	NOISE
codedb_hot	102501	98519	-3.88%	-3982	OK
codedb_outline	245858	251344	+2.23%	+5486	OK
codedb_read	87445	94395	+7.95%	+6950	OK
codedb_search	184574	176955	-4.13%	-7619	OK
codedb_snapshot	2835681	1351605	-52.34%	-1484076	OK
codedb_status	210677	212565	+0.90%	+1888	OK
codedb_symbol	59665	59284	-0.64%	-381	OK
codedb_tree	70447	79217	+12.45%	+8770	NOISE
codedb_word	71140	69147	-2.80%	-1993	OK

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f6f12b6c40

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Rebuild symbol index from outlines before snapshot emit

This change serializes symbol_index directly from explorer.symbol_index, but that map is not guaranteed complete after fast snapshot restore: insertRestoredFile (src/snapshot.zig) restores outlines/content without calling rebuildSymbolIndexFor, and findAllSymbols explicitly documents that symbol_index can be incomplete after this path (src/explore.zig, comment around lines 1325-1328). In warm-start loads, many unchanged files follow that restore path, so buildSnapshot now returns a partial or empty symbol_index even when outlines are present. Please retain an outline-based fallback (as before) when emitting snapshot symbol data.

Useful? React with 👍 / 👎.