Loadable lifecycle improvements: abort handling and unified exception hook

[lbwexler]

Builds on spanItAll. Reduces per-model boilerplate and makes handling of stale, obsolete, and auto-refresh load errors consistent across the Loadable lifecycle.

Additions

• LoadAbortedException (isRoutine + isAborted) - thrown when a load is superseded. Raised automatically by FetchService after each fetch (and on demand via LoadSpec.abortIfNeeded() for non-fetch async work), based on LoadSpec.shouldAbort. Recognized by ExceptionHandler and dropped silently.
• Loadable.skipStaleLoads (default true) - skip/abort loads superseded by a newer started request. Loads superseded by a newer completed request are always skipped.
• Loadable.skipAutoRefreshErrors (default true) - silence errors raised during an auto-refresh.
• Loadable.handleLoadException(e, loadSpec) - new hook called only for surface-worthy failures (abort exceptions and, per the flags above, stale/auto-refresh errors are filtered out first). Default delegates to XH.handleException(e). Now a required member of the Loadable interface; HoistModel/HoistService provide the default.

Other changes

• Renamed the isFetchAborted exception flag to isAborted (now also set by LoadAbortedException). Breaking - update references in catchWhen/catchDefaultWhen predicates etc.
• Fixed LoadSpec.isStale to compare loadNumber (stable across cloneWithSpan) rather than object identity.
• Migrated the admin/cmp models off hand-rolled if (loadSpec.isStale) return / .catch boilerplate to the new hook.

Behavior change to default exception handling — open for discussion

Previously, LoadSupport caught an exception from doLoadAsync only to record it and then re-threw it; it never called XH.handleException itself. For a surface-worthy load that an app did not handle, the error therefore reached the user only if something downstream surfaced it — and the standard refresh lifecycle (loadAllAsync via RefreshContextModel) swallows rejections with Promise.allSettled + logError, so the common outcome was a console log, no dialog.

Now LoadSupport routes such errors to handleLoadException (default: XH.handleException(e)), so a non-auto-refresh load that throws and isn't handled by the app will surface a standard error dialog by default. (Superseded/auto-refresh failures remain silent — no alert; genuine non-abort errors are still logged server-side.) Note also that loadAsync() no longer rejects for these errors — it resolves and routes them to the hook.

This is arguably the better default (a failed initial load surfaces instead of leaving a blank view), but it is an observable change for apps that relied on the old console-only behavior. If we want to soften the default, we could scale it back to either or both of:

• a) Log on server only - default handleLoadException calls XH.handleException(e, {showAlert: false}), so failures are recorded but not shown to the user.
• b) Re-throw - have the default re-throw (and/or have loadAsync() continue to reject), more closely matching the current behavior so existing callers/lifecycle handling stay in control.