[aheejin]

In Emscripten EH, #25988 changed where we increment refcount to fix a refcounting inconsistency between Emscripten vs. Wasm EH (#17115). Before #25988, it was incremented in __cxa_begin_catch, but after
#25988, it was in __cxa_throw and __cxa_rethrow, more inline with
Wasm EH.

__cxa_end_catch calls __cxa_decrement_exception_refcount, which decrements the refcount and destroys/frees the exception. But when an exception is rethrown, we should not destroy it in __cxa_end_catch. The below illutrates how a rethrown exception is (correctly) not destroyed in __cxa_end_catch:

// Before #25988 (refcount incremented in __cxa_begin_catch):
try {
  throw MyEx(); // __cxa_throw:       refcount = 0, rethrown = false
} catch (...) { // __cxa_begin_catch: refcount = 1 (0 -> 1), rethrown = false
  throw;        // __cxa_rethrow:     refcount = 1 (unchanged), rethrown = true
}               // __cxa_end_catch:   refcount = 0 (1 -> 0); not destroyed/freed because rethrown == true

// After #25988 (refcount incremented in __cxa_throw and __cxa_rethrow):
try {
  throw MyEx(); // __cxa_throw:       refcount = 1 (0 -> 1)
} catch (...) { // __cxa_begin_catch: refcount = 1 (unchanged)
  throw;        // __cxa_rethrow:     refcount = 2 (1 -> 2)
}               // __cxa_end_catch:   refcount = 1 (2 -> 1); not destroyed/freed because refcount > 0

So after #25988 we were supposed to solely rely on the refcount to determine whether to destroy an exception. But the !exception_header->rethrown check was incorrectly still left in __cxa_decrement_exception_refcount. This didn't affect C++-only programs, but this prevented an exception from being destroyed when an exception was rethrown to JS and decrementExceptionRefcount was called from there, as in the test added here.

This removes the incorrect rethrown flag check. After this and #26518, no code uses the rethrown field, so this also removes it and its related methods.

This is a part of fixes for #27705.