fix(installer): honor CODEDB_NO_HOOKS + removal marker so hook removal sticks (#658)

[justrach]

The remaining half of #658 (problem c): the installer silently re-added the codedb-block-legacy.sh PreToolUse hook on every run, even after the user removed it (reporter removed it 3×). The scope+redirect half was #663; this makes removal stick.

• CODEDB_NO_HOOKS=1 at install time skips the PreToolUse hook merge entirely (SessionStart warmup unchanged), and writes a ~/.codedb/no-hooks marker so the opt-out persists across future installs without re-passing the env var.
• If the marker exists, the merge is skipped regardless of env.
• Default behavior (no env, no marker) is unchanged.

Verified: installer Python heredoc compiles; bash -n OK; simulated all three scenarios (opt-out creates marker + skips; marker-present later run still skips; baseline unchanged). Produced by a Sonnet agent; independently re-verified.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3eb1a0dccc

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]

Remove existing hook when opt-out is active

When CODEDB_NO_HOOKS=1 is used on a machine that already has the codedb PreToolUse entry in ~/.claude/settings.json, this branch only skips calling merge_hook; it never deletes the existing entry, and the generated hook script only checks the runtime env var, not the new ~/.codedb/no-hooks marker. As a result, the install-time opt-out persists the marker but the legacy-tool blocker still runs in later Claude sessions that do not export CODEDB_NO_HOOKS, so existing users cannot actually disable the hook via the new installer path.

Useful? React with 👍 / 👎.

[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	67615	68281	+0.98%	+666	OK
codedb_changes	12077	12818	+6.14%	+741	OK
codedb_context	310652	312455	+0.58%	+1803	OK
codedb_deps	345	383	+11.01%	+38	NOISE
codedb_edit	41074	41828	+1.84%	+754	OK
codedb_find	3064	3051	-0.42%	-13	OK
codedb_hot	27910	29132	+4.38%	+1222	OK
codedb_outline	18456	17607	-4.60%	-849	OK
codedb_read	14280	15023	+5.20%	+743	OK
codedb_search	65187	69004	+5.86%	+3817	OK
codedb_snapshot	75939	74911	-1.35%	-1028	OK
codedb_status	11708	11285	-3.61%	-423	OK
codedb_symbol	56971	54804	-3.80%	-2167	OK
codedb_tree	23119	21869	-5.41%	-1250	OK
codedb_word	12825	12705	-0.94%	-120	OK