Add presets — named, switchable configurations

[engineer-myoa]

Proposed and discussed first in #1818 — posting the PR now since you said you'd be inclined to move forward.

I switch between a full-size keyboard, a tenkeyless one, and the MacBook's built-in keyboard. Most of my shortcuts sit on the numeric keypad, so on the other two they're unreachable. Import/Export can move a configuration around, but it's two file dialogs each time, there's no notion of a named set, and a config file can't express "this action is deliberately unbound" — Defaults.encoded() skips the empty-dictionary state, so with #1805 making import authoritative the key is removed and the built-in default comes back instead.

What this adds

A preset is a named snapshot of the shortcuts plus the settings that describe window behavior. Settings about the app itself — launch at login, update checks, menu bar icon, per-app exclusions, transient Todo state — stay global. The exclusion list is explicit and tested against Defaults.array so a future rename can't silently pull a key in.

Shortcuts are stored in all three states MASShortcut distinguishes (absent / explicitly unbound / assigned), so switching restores unbound bindings as unbound. Snapshots read persistentDomain(forName:) rather than dictionary(forKey:) to tell an untouched shortcut from an assigned one.

The active preset mirrors the live settings, so there's no save button or dirty state. Applying one writes settings first (so alternateDefaultShortcuts is current before registerDefaults() re-runs), then shortcuts, then posts configImported and lets the existing observers reconcile. Presets can be created from the built-in defaults or by duplicating the current one, and renamed or deleted.

UI

Per your feedback, the picker is in the Extras popover under a "Presets" header rather than the Shortcuts tab. No storyboard changes. PrefsViewController keeps only a configImported observer so the shortcut recorders reflect a newly applied preset.

That observer is registered in viewDidLoad rather than awakeFromNib, because awakeFromNib runs twice for that scene — everything already in it happens to be idempotent, so it hadn't surfaced before.

Export format

Presets ride along in Import/Export through a new optional Config.presets field rather than another entry in defaults, which would have put the whole library into the file as one escaped JSON string. Both directions stay compatible and there are tests for it: builds without the field ignore it when decoding, and the field is optional so older configs still load. Config keeps a memberwise initializer with presets defaulting to nil, so existing call sites are unchanged.

This is the one place the feature steps outside the "every new key goes in Defaults.array" convention. The convention's purpose — never silently dropping a key from export — is still met by the dedicated field.

Along the way

• Defaults.load(fileUrl:)'s settings loop is extracted to Defaults.apply(defaults:) so import and preset switching share a path. Pure extraction, no behavior change.
• JSONDefault.load(from:) now clears its decoded value when a config sets the underlying string to nil; previously footprint colors and snap areas couldn't be reset through import. Different path from the init(key:defaultValue:) fix in Honor defaults-write customizations for JSONDefault keys #1811.

I've kept an eye on #1782 since you mentioned it — the Defaults.apply(defaults:) seam is small and isolated, so it shouldn't get in the way of reworking configuration management later.

Testing

29 new tests, all passing. For reference, f6279e2 runs 223 tests with 22 failing assertions across ActiveSideSplitRatiosCooperativeTests, CooperativeCornerResizeTests and HalfSplitCornerCalculationTests; this branch runs 252 with the same 22. I checked an unmodified checkout to confirm they aren't from this work — happy to open a separate issue for them if useful.

Happy to adjust anything, including hiding the picker behind a preference if you'd rather it not be visible by default.

🤖 Generated with Claude Code

[rxhanson]

Thanks for following through with contributing! There's a decent amount to unpack and test out here, so it'll take me a little bit to walk through this one. I'm planning on cutting a release in the next day or so, and will dive a little deeper into this after the release.