fix(cli): key combos survive NumLock/CapsLock under kitty CSI-u (salvage #89676 + #90291)

[kshitijk4poor]

Summary

Key combos no longer die (leak as literal text like [99;133u) when NumLock or CapsLock is on under the kitty keyboard protocol. Closes #88221, closes #89651.

kitty/ghostty OR lock-key state into the modifier parameter of every key event while a lock is on (CapsLock=+64, NumLock=+128, both=+192), so Ctrl+C arrives as ESC[99;133u instead of ESC[99;5u and plain Down as ESC[1;129B instead of ESC[B. None of those variants were in ANSI_SEQUENCES, so with a lock enabled every combo — and on keypads every plain nav key — broke.

Composite salvage of the two strongest PRs in the 4-PR cluster, with authorship preserved via cherry-pick:

• fix(cli): map kitty CSI-u lock-bit variants so key combos survive NumLock #89676 (@liuhao1024) — CSI-u lock twins in _install_paired + the Esc-key family, with the correct producer-contract insight that the xterm modifyOtherKeys ESC[27;N;CP~ form never carries lock bits (so it gets no twins), plus a strong test suite.
• fix(cli): map NumLock/CapsLock modifier-bit variants under kitty protocol #90291 (@grahfmusic) — the legacy CSI-letter/CSI-tilde navigation slice (ESC[1;129B arrows, ESC[3;129~ Delete, …) that every CSI-u-focused PR missed.

Changes

• hermes_cli/pt_input_extras.py:
  • shared _lock_variants() helper; both cherry-picks and all follow-up sites use it
  • install_shift_enter_alias / install_ctrl_enter_alias / install_cmd_backspace_alias: CSI-u spellings get lock twins (Shift+Enter with NumLock = ESC[13;130u previously fell through)
  • legacy nav lock twins now derived from the existing table for all modifiers 1–16 (not just plain/shift as in fix(cli): map NumLock/CapsLock modifier-bit variants under kitty protocol #90291), covering Ctrl/Alt/Cmd-modified arrows + F1–F4 (SS3 base fallback) under locks
  • unmodified CSI-u keys (Tab/Enter/Space/Backspace at modifier 1) + lock twins
  • kitty PUA functional keys (keypad, F13–F24, Ignore range) + lock twins — NumLock especially matters here since it gates the keypad itself
• tests/cli/test_modify_other_keys_aliases.py: contributors' tests + 8 follow-up tests for the widened surface
• contributors/emails/: mapping for @grahfmusic

Validation

	Before	After
Ctrl+C + NumLock (ESC[99;133u)	literal text leak	Keys.ControlC
plain Down + NumLock (ESC[1;129B)	literal text leak	Keys.Down
Shift+Enter + NumLock (ESC[13;130u)	literal text leak	newline
Cmd+Backspace + NumLock (ESC[127;137u)	literal text leak	kill-line
KP_0 + NumLock (ESC[57399;129u)	literal text leak	"0"
modifyOtherKeys tilde form	no lock twins	still no lock twins (negative-tested)

• 213 passed in tests/cli/test_modify_other_keys_aliases.py, 216+2s across tests/cli/
• Byte-stream E2E through a real Vt100Parser: 8-sequence locked stream parses with zero CSI leaks
• Idempotency: second install pass registers 0 sequences
• Behavior parity vs origin/main on 14 non-lock sequences: identical

Credit

• @liuhao1024 — fix(cli): map kitty CSI-u lock-bit variants so key combos survive NumLock #89676 (cherry-picked, authorship preserved)
• @grahfmusic — fix(cli): map NumLock/CapsLock modifier-bit variants under kitty protocol #90291 (cherry-picked, authorship preserved)
• @Schrotti77 (fix(cli): decode kitty CSI-u lock-key modifier bits (NumLock/CapsLock) #88265) and @iamlukethedev (fix(cli): map kitty CSI-u sequences with the num_lock modifier bit #88673) independently diagnosed the same root cause; their PRs overlap the salvaged coverage and will be closed with credit.