fix(cli): make Shift+letter actually type a capital under modifyOtherKeys

[MwC-Trexx]

Summary

• Patch prompt_toolkit's VT100 parser so character-valued ANSI_SEQUENCES entries insert the character, not the matched escape bytes
• Preserve intentionally blanked insert_text on multi-key ANSI tails (Alt / Shift+Alt must not type the letter)
• Add PromptSession-level regression tests for Shift+letter, Shift+Space, Alt, and Shift+Alt under modifyOtherKeys level 2 / CSI-u

Motivation

#87511 mapped ESC[27;2;<code>~ / ESC[<code>;2u to uppercase letters, but the classic CLI still leaks raw CSI into the buffer on Ghostty (and iTerm2 / WezTerm / kitty). prompt_toolkit builds KeyPress(key=<table value>, data=<matched bytes>) and the default Keys.Any binding inserts event.data — so mapping fixed what the key is, not what it types.

Verified on current main before this change: Shift+W then hat submits \x1b[27;2;87~hat. After: What.

Changes

• hermes_cli/pt_input_extras.py — _install_literal_key_data_patch() (defensive getattr, marker on wrapper, only rewrites non-empty insert_text != key, called at end of install_modify_other_keys_aliases())
• tests/cli/test_modify_other_keys_insert_text.py — asserts submitted text via real PromptSession + pipe input; includes Alt/Shift+Alt non-insert guards

Related

• Fixes the remaining half of Classic CLI: Shift+letter prints literal [27;2;<code>~ instead of capital on Ghostty (modifyOtherKeys level 2) #87390 / [Bug] Unmapped xterm modifyOtherKeys / CSI-u sequences insert literal text into the prompt (Shift+Space, Ctrl+K, …) #86866 after fix(cli): map all Ctrl/Alt/Shift+key combos under modifyOtherKeys level 2 #87511
• Refs fix(cli): make Shift+letter actually type a capital under modifyOtherKeys #87785 — same problem class (krunkosaurus); this is a rebased approach on current main that also keeps multi-key tail blanks intact. Happy to coordinate with that PR rather than race it.
• Orthogonal to fix(cli): map Shift+letter for all cased scripts under modifyOtherKeys #87637 (non-Latin Shift map extension) — that PR extends the table and inherits the same insert-text gap unless something like this lands

Test plan

• ./venv/bin/python -m pytest tests/cli/test_modify_other_keys_insert_text.py tests/cli/test_modify_other_keys_aliases.py -q — 233 passed
• Pipe repro: \x1b[27;2;87~hat \x1b[27;2;72~appens → What Happens
• Alt+a / Shift+Alt+a do not insert into the buffer
• Manual on Ghostty classic CLI: type capitals with Shift — no [27;2;…~ leak; Shift+Enter still newline; Alt chords unchanged

Notes for reviewers

• No platform-specific code; pure prompt_toolkit layer
• Keys is a str Enum — not isinstance(key, Keys) is load-bearing even though every current Keys value is multi-char
• Ctrl/Alt Keys-valued entries, named keys, bracketed paste, and Shift+Enter aliases are intentionally unchanged
• Private _call_handler is patched defensively (getattr + try); missing/renamed API degrades to no-op without breaking later installers

[alt-glitch]

This was generated by AI during triage.

Related to #87785: both repair the same prompt_toolkit self-insert path. This revision explicitly preserves blank insert_text on multi-key Alt and Shift+Alt tails; maintainers should choose or consolidate the two approaches.

[MwC-Trexx]

Thanks for taking a look.

Pushed a small follow-up on this same PR (no new PR):

• Keep multi-key ANSI tails blanked (insert_text only rewritten when non-empty and different from the key), so Alt / Shift+Alt do not type the letter
• Soften the docstring around “what changes shape”
• Note that Keys is a str Enum, so not isinstance(key, Keys) is intentional and load-bearing
• Regression tests for Alt and Shift+Alt non-insert

tests/cli/test_modify_other_keys_insert_text.py + tests/cli/test_modify_other_keys_aliases.py: 233 passed.

Happy to coordinate with #87785 if that is preferred as the landing vehicle.