docs: record five user-facing changes that merged without an entry

[xxx]

The changelog convention is that an entry lands with its PR. Auditing all 101 commits since v1.1.0 for whether they touched CHANGELOG.md found five that changed what a user sees and carried no entry.

PR	What a user would notice
#354	A native Android click or set_value through the companion dispatched by glass's own element number, which diverges from the device's numbering past the first pruned subtree — so on any truncated tree the action reached a different element and reported success.
#353	A device failure was classified by matching words in the rendered message, whose tail is the device's own output, so a wedged emulator could be read as AccessibilityNotReady — the condition glass_wait_for_element deliberately waits out.
#330	An element whose screen had been destroyed was reported as changed since the snapshot, advising a re-snapshot that cannot help.
#310	A subtree dropped by a failed child read left the tree looking complete, indistinguishable from one where nothing was there.
#283	glass_clipboard_set on the iOS Simulator returned ok for a payload still in flight.

All five correct what glass reports rather than adding to it, so they join ### Fixed, prepended in the section's existing newest-first order.

The other commits with no entry are ci, test, refactor, chore, or internal docs, which the convention excludes. Two judgement calls worth naming: #311 (the WGC probe moved to a thread this code owns) fixed a crash in the test binary and changes nothing glass doctor reports, and the edition-2024 and rmcp/tonic bumps change no tool surface — none gets an entry.

No code changes.

Verification

cargo test --workspace — 1938 passed, 0 failed. cargo fmt --check clean.

🤖 Generated with Claude Code