chore: release v0.11.0 — the stability candidate

[vyncint]

The release commit for 0.11.0, the stability candidate, cut per docs/RELEASING.md steps 1–3 (step 2b included):

• workspace.package.version → 0.11.0; termlens-cli's termlens = { version = "0.11.0", … }; Cargo.lock refreshed — the diff is nine member version lines and nothing else.
• [Unreleased] → [0.11.0] - 2026-09-11, a fresh empty [Unreleased] above. The section opens with the one breaking change (Screen::unsupported() exposes its Arc<str> storage and keeps the overflow count in a second accessor #330) and its migration table; .github/scripts/extract-changelog.sh v0.11.0 extracts 179 lines of notes.
• crates/termlens/tests/compat/0.11.0/ frozen from this tree with the write_corpus writer: six text files byte-identical to 0.10.1's, JSON with "format": 1. The compat README says how later directories come to be.
• The report action tag in the README and the skill → @v0.11.0 (exists once the tag is pushed).

The gate on this PR: tree 0.11.0 against published 0.10.3; check-semver.sh reads the - **Breaking:** bullet under [0.11.0] (ahead of the baseline) and runs with --release-type major — measured locally: PASS in all three views. The same holds for the tag's re-run of ci.yml. After publish, a follow-up PR moves the baseline to 0.11.0 (RELEASING step 5).

Local matrix on this exact tree (all green): fmt; clippy all-features / default / minimal / decode / Windows target; rustdoc both ways; cargo test --workspace --all-features 529 passed, 0 failed, 5 ignored (each ignore platform-reasoned or the corpus writer); default set; the four isolated -p termlens sets; feature isolation; CLI contract; README links; skill snippets; gates-listed (19); candidate statement; semver self-test; cargo-deny; MSRV 1.85.1 over three configurations; cargo package -p termlens --locked verified (305 KB, contents listed and reviewed — corpus and snapshots ship as tests, nothing private). cargo package -p termlens-cli cannot resolve termlens 0.11.0 until the library is published, as in every prior release; release.yml publishes the library first.

Performance (release builds, same input, interleaved, new first): JSON serialize 228 vs 235 µs, deserialize 432 vs 432 µs, with_styles 29.5 vs 30.3 µs (0.11 vs 0.10.3); 40k-line flood through 80x24 within noise. No regression; the JSON gains 11 bytes for the format field.

Stress: run 34527017048 on the candidate tree is in flight; per RELEASING step 0 the stress workflow also runs on the merged release commit before the tag is pushed.

After merge: wait for stress on the exact commit → tag v0.11.0 → release.yml → verify crates.io, the GitHub Release, install.yml, a fresh consumer → baseline PR → consumer migrations (separate repositories, against crates.io).