fix(web): decide the composer strip from the controls it will draw

[rynfar]

Closes #435. Substantially addresses #437.

The composer context strip decided its own visibility from isGitRepo, while
BranchToolbar decided its Git controls from isGitRepo && !rollbackActive.
During a rollback the two disagreed: the strip stayed open because there was a
repository, and then drew nothing, leaving a visible empty bar with the shell's
clip-path extension around it.

The predicate's input was named for the repository when what it needs is what
the strip will actually draw, so it is now showsGitControls, and ChatView
derives that once and passes the same value to both the predicate and
BranchToolbar. The two can no longer disagree.

While here, the mount-versus-visible pair moved out of the view into
resolveComposerContextStripVisibility in ChatView.logic.ts. That pair is the
part that has now been wrong twice — once hiding a live capacity reading (#432),
once showing an empty bar — and it was only reachable by rendering a chat. It is
now a pure function with tests covering the rollback case, the capacity case,
and the measuring strip that must stay mounted while hidden.

Not claiming #437 is closed: ChatView still assembles the inputs, so cutting a
field there would restore a bug with tests green. What is left is literal field
assignment rather than a decision, and covering it properly needs a ChatView
render harness, which does not exist yet.

No UI change to show: this fixes a state that renders nothing either way — an
empty bar becomes no bar. Verified by test rather than by screenshot, including
a mutation check (forcing the Git input true fails 3 tests).

Written by Claude Opus 5 in Claude Code.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.8 KiB	13.8 KiB	−12 B (−0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.7 KiB	6.7 KiB	−11 B (−0.2%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	57.4 KiB	57.4 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.8 KiB	13.8 KiB	−4 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	−8 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.7 KiB	6.7 KiB	+4 B (+0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.2 KiB	58.2 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: f669dcd · PR result: 9cf70fd · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 111.2 KiB
• Claude decoded thread snapshot: 111.9 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.