Fix __cxa_find_matching_catch's memory leak

[aheejin]

__cxa_find_matching_catch allocated a buffer of size 4 using malloc only
when it was first called and reused it through the entire program run. And this
malloc'ed memory wasn't freed. Because this is only allocated once, I changed it
to a static allocation using makeStaticAlloc.

Fixes #8919.

[kripken]

Nice!

This can be slightly simpler: a new global variable (__cxa_find_matching_catch_buffer) is not needed, as you can write

var buffer = {{{ makeStaticAlloc(4) }}};

inside the one function that uses that buffer (the {{{ .. }}} will be expanded out into the absolute address in memory). lgtm with that.

[kripken]

Hmm that CI error on test_clear_error_on_massive_static_data is worrying. There should be no change to static memory from this PR, but I think that error means that there is.

I think the issue is that we include that JS file based on DISABLE_EXCEPTION_THROWING and not _CATCHING. I don't remember all the details for the difference (some notes are in settings.js). But I think we do need to ifdef this on DISABLE_EXCEPTION_CATCHING (which is on by default, unlike the other one).

[aheejin]

Hmm that CI error on test_clear_error_on_massive_static_data is worrying. There should be no change to static memory from this PR, but I think that error means that there is.

I think the issue is that we include that JS file based on DISABLE_EXCEPTION_THROWING and not _CATCHING. I don't remember all the details for the difference (some notes are in settings.js). But I think we do need to ifdef this on DISABLE_EXCEPTION_CATCHING (which is on by default, unlike the other one).

test_clear_error_on_massive_static_data works on my local machine, so it's weird. I addressed feedback and merged incoming just now, so I'll wait for the CI to complete and see if that happens again.