[Sema]: minor improvements to unused expression result diagnostics

[jamieQ]

Previously we would emit a diagnostic in cases in which a throwing function returning a structurally-uninhabited type was called within an optional try expression. Such cases can never produce a result other than nil, so the existing diagnostic seemed not particularly helpful. This change attempts to address this and other minor inconsistencies by expanding the isDiscardableType() logic to encompass "structurally uninhabited" types (e.g. tuples that eventually bottom out in an uninhabited type in some position), and also look through any levels of optionality to find such types. This makes the unused expression result diagnostics a bit more consistent in that they will now allow 'uninteresting' unused results to be implicitly discarded. In particular, the following cases now have the same diagnostic behavior (namely, do not produce a warning for the unused results):

func maybeNever() -> Never? { nil }
func maybeVoid() -> Void? { nil }

maybeNever() // used to warn with: 'warning: result of call to 'maybeNever()' is unused'
maybeVoid()

Resolves: #85092

[jamieQ]

@swift-ci please smoke test

[hamishknight]

Thank you! This looks reasonable to me

[grynspan]

Looks good to me. Is there a practical difference between isUninhabited() and isStructurallyUninhabited() that we need to consider?

[xedin]

It has to do with tuples, I think what we have in the PR is reasonable.