[Wildcard Variables] Update `UNUSED_CATCH_CLAUSE` reporting

[pq]

Like UNUSED_LOCAL_VARIABLEs (#55719), UNUSED_CATCH_CLAUSE should not special case __s.

This adjustment is straight-forward in the presence of an on clause where __s are special cased, however the case without an on is interesting... Currently we do not report unused catch clauses if there is no on clause regardless of underscores.

For example:

try {
} catch(e) { }

is treated the same as

try {
} catch(_) { }

(And neither produce a diagnostic.)

Question: with wildcard variables, should we start reporting UNUSED_CATCH_CLAUSE on all bound variables (and so on the catch(e) case above)?

(Note that this will be a bit disruptive -- luckily the quick fix is easy to implement.)

/cc @kallentu @bwilkerson @munificent

[lrhn]

Since you have to write something to catch an error, either } on Object { or } catch (_) {, we should probably not warn about the latter.

We could warn about one of on Object catch (_) being unnecessary. Probably the latter, we also have a lint against not having an on clause.
I wouldn't say anything about } catch (_) {, } catch (_, _) { or even } catch (e, _) {. They're fine as they are. Slightly longer than necessary for the last two, but not something worth bothering the author about unless they have a specific "As short as possible" lint.

[kallentu]

I talked about this briefly with @pq today.

From what we discussed, there's some history with allowing exception variables (the e) to be unused. Even the lint examples for empty catch blocks show an un-opinionated example of catch clauses being ignored.

Maybe this is the right time, now that we have a mechanism for wildcards to push users to using _ if they're not actually using the exception variable?
Having the edge case where the exception variable can be unused in catches (with no on) seems weird, especially in a world where wildcards exist.

It'll be breaking, but the fix for it is simple (e -> _).

[pq]

Maybe this is the right time, now that we have a mechanism for wildcards to push users to using _ if they're not actually using the exception variable?

💯

See #55719 for a proposal to start flagging all unused (non-wildcard) exception variables as UNUSED_LOCAL_VARIABLEs.