refactor(gui): split keyboard-shortcut system into MainWindow_Shortcuts.cpp (#3351 Phase 1c)

[ten9876]

Summary

Fourth PR of the #3351 series. Pure code motion — no behavior change, no MainWindow.h changes.

Moves the complete keyboard-shortcut system into the new MainWindow_Shortcuts.cpp (1,087 lines):

Piece	Lines
Shortcut-state definitions + accessors (s_keyboardShortcutsEnabled, shortcutGuard, …)	~40
registerShortcutActions() — the full action table	559
Slider shortcut lease (#745): begin/renew/release + eventFilter()	~365
handleCwMomentaryShortcut() + trivial keyPress/keyRelease overrides	~65

This completes the handoff MainWindowShortcutState.h promised in Phase 1b — declarations there, definitions now owned by the Shortcuts TU.

Not-pure-motion exceptions (the review surface)

1. kSliderShortcutLeaseMs moved with its only user.
2. panCountForLayoutId promoted from MainWindow.cpp's anon namespace to MainWindowHelpers.{h,cpp} — registerShortcutActions (moved) plus two MainWindow.cpp call sites now share it cross-TU. Pure map lookup, fits the helpers file's no-state rule.
3. isCwMomentaryActionId deliberately NOT moved — its remaining caller is the MIDI dispatch path still in MainWindow.cpp.

Everything else verbatim. Breadcrumbs at all six excision sites.

Numbers

• MainWindow.cpp: 16,261 → 15,239 (−1,022)
• Series running total: 19,474 → 15,239 (−21.7%)

Verification

• Full build clean (Linux x86, RelWithDebInfo)
• All 33 ctest suites pass (incl. the new radio_discovery_test from main)
• a11y lint: 0 findings

QA checklist for the bench

The shortcut system touches every keyboard interaction:

• Global shortcuts fire when enabled (try a handful from Help → Shortcuts: band up/down, mode, MOX)
• View-menu shortcut toggle still gates them (off = dead, on = live)
• Typing in a text field (memory name, frequency entry) does NOT trigger shortcuts
• Click a slider → arrow keys nudge the slider (lease), then ~2 s after release, arrows return to global shortcuts
• Same lease behavior on a TCI/DAX MeterSlider fader
• CW momentary keys (straight key / paddles) press-and-hold from keyboard if you have them bound
• App quit via window close while connected — clean shutdown (eventFilter intercepts Quit)

Refs #3351

[aethersdr-agent]

Reviewed with a mechanical pure-motion check: I diffed every line removed from MainWindow.cpp against the new MainWindow_Shortcuts.cpp (whitespace-normalized). Every removed line is accounted for verbatim — the only lines not landing in the Shortcuts TU are the panCountForLayoutId body, which appears identically in MainWindowHelpers.cpp. Conversely, the only content in the new file that wasn't moved is the header comment, includes, and namespace scaffolding. The "pure code motion" claim holds.

Cross-TU wiring also checks out:

• The shortcut state (s_keyboardShortcutsEnabled, s_sliderShortcutLeaseActive) and its accessors now have exactly one definition site (the Shortcuts TU), matching the extern declarations in MainWindowShortcutState.h from Phase 1b. The only outside user is MainWindow_Menus.cpp (View-menu toggle + shortcutGuard), which is the design that header documents.
• isCwMomentaryActionId correctly stays behind — its remaining caller is the MIDI dispatch path at MainWindow.cpp:3749.
• panCountForLayoutId in the helpers file is a stateless map lookup, consistent with that file's no-state rule, and both remaining MainWindow.cpp call sites plus the moved eventFilter use the shared helper.
• All six moved definitions (eventFilter, registerShortcutActions, the three lease functions, handleCwMomentaryShortcut, key event overrides) exist exactly once across the GUI sources. CMake entry added; no MainWindow.h changes, as advertised.

Two minor notes, neither blocking:

1. Missing <algorithm>/<iterator> includes. The new TU uses std::clamp, std::min/std::max, and std::size (in registerShortcutActions) but includes neither header — it currently compiles everywhere (all CI green, including MSVC) via transitive includes only. Both MainWindow.cpp and the Phase 1b sibling MainWindow_Menus.cpp include <algorithm> explicitly; suggest adding <algorithm> and <iterator> here to match and to avoid a fragile dependency on Qt header internals across upgrades.
2. Optional: MainWindow.cpp still includes MainWindowShortcutState.h but no longer references any of its symbols after this move — the include can be dropped (fine to leave if a later phase will need it again).

Nice work — the breadcrumbs at the excision sites, the verbatim motion, and the explicit not-pure-motion list made this straightforward to verify. The QA checklist for the bench is appreciated; the lease and CW momentary paths are exactly the ones worth a hands-on pass.

---

_{🤖 aethersdr-agent · cost: $17.2202 · model: claude-fable-5}

[ten9876]

Both nits addressed in `2670835b`:

1. `` + `` added to MainWindow_Shortcuts.cpp. Verified the usages you flagged (`std::size` at the step tables, `std::clamp` on RF gain, `std::min`/`max` on audio gain) — all were riding transitive Qt includes. Now explicit, matching MainWindow.cpp and MainWindow_Menus.cpp.

2. Dead `MainWindowShortcutState.h` include dropped from MainWindow.cpp. Confirmed its only remaining "references" were the two breadcrumb comments. Took the drop-it-now option — if a later phase needs the state again, that phase re-adds the include alongside its actual user (IWYU over speculative retention).

Thanks for the mechanical pure-motion verification — diffing removed-vs-added lines whitespace-normalized is exactly the right review for this series, and it's the check I want every phase to pass.

All 33 ctest suites still green after the fixes.

[jensenpat]

Reviewed: verified pure code motion — every line removed from MainWindow.cpp reappears verbatim in MainWindow_Shortcuts.cpp/MainWindowHelpers; only new lines are includes, breadcrumbs, and the panCountForLayoutId declaration. All CI green.