Emit declarations using alias chains

[Andarist]

fixes #57415

[weswigham]

Obligatory: Is there a linked issue?

AFAIK this is sensible, but probably expensive - getAliasForSymbolInContainer is already costly, since it's an exhaustive traversal of all symbols in scope looking for aliases that refer to the target symbol - making it recursive is going to increase the worst case cost exponentially without some very smart caching. And, perhaps moreover, getAliasForSymbolInContainer is basically just a duplicated tiny slice of the inner loop of getAccessibleSymbolChain without all the recursion, safety, or caching rigamarole, to try and prevent the need for the (still expensive) getAccessibleSymbolChain call. By making it return a chain like this, isn't it basically just a copy of getAccessibleSymbolChain that preferences aliases, without all the aforementioned safety valves? Maybe it'd just be better to modify the useExternalAliasingOnly parameter into an alias-preference-flags parameter so getAccessibleSymbolChain is suitable for use here directly?

I say that, but I'm also actively working on a replacement for getAccessibleSymbolChain itself that's a bit less.... bad. Exhaustive scope traversal on lookup is not a good thing (there are many scopes with many symbols), and I'd like to remove it; while we don't have the perf numbers for it, getAccessibleSymbolChain has basically organically grown in function since TS 1.8 or earlier, making it the oldest part of the declaration emit pipeline that's never seen a major refactor, and its' implementation is pretty unoptimized (but with 5+ years of edge cases written into it, hgggg) - it's come up as a hotspot in specific examples multiple times in the past, but we've always worked around the issue by adding limiters higher up in the declaration emit or checking process.

I'd also say "we want to check the perf on this, at least", but declaration emit performance is still a bit of a blind spot in our perf suite. It's usually only lightly exercised by the error messages generated by the perf tests. 😭

[weswigham]

Yeah, definitely see if this can be integrated into the normal getAccessibleSymbolChain function. It has some inefficiencies that I'm looking into fixing, but so does this solution as-written, so better to keep them in one place.

[fubhy]

Also linked (atm. unclear if this fixes it though): #55021