This builds on the existing work to make assertion thread safe,
by adding an extra synchronization point in the holder of
`ReusableStringStream`'s stream instances, as those are used to
build the messages, and finishing the move of message scope holders
to be thread-local.

As with the JSON writer, the old code was made to be simple and
for each char just decided whether it needs escaping, or should be
written as-is. The new code instead looks for characters that need
escaping and batches writes of characters that do not.

This provides 4-8x speedup (length dependent) for writing strings
that do not need escaping, and keeps roughly the same performance
for those that do need escaping.

The old code was exceedingly simple, as it went char-by-char and
decided whether to write it to the output stream as-is, or escaped.
This caused a _lot_ of stream writes of individual characters.

The new code instead looks for characters that need escaping, and
bulk-writes the non-escaped characters in between them. This leads
to about the same performance for strings that comprise of only
escaped characters, and 3-10x improvement for strings without any
escaping needed.

In practice, we should expect the former rather than the latter,
but this is still nice improvement.

Primarily this means that with Clang we use `__builtin_debugtrap`
if available.

As it turns out, enums can be used to declare bitfields, and we
cannot form a reference to a bitfield in the cpature. Thus, we add
`std::is_enum` as a criteria to the default for `capture_by_value`,
so that enum-based bitfields are also captured by value and thus
decomposable.

Closes #3001

Having `std::terminate` as the backstop after `__assume(false)`
would trigger W4702 (unreachable code) with MSVC. As we want to
keep `__assume(false)` for the optimization hint in NDEBUG builds,
we have to avoid mixing in `std::terminate` for those builds.

Fixes #3007

Bad handling of `TestFailureException` when translating unexpected
exceptions inside assertion macros led to the unexpected exceptions
handling erroring out through throwing the same exception again.

This was then backstopped by the machinery for handling uncaught
exceptions from assertions, which is normally used by the
`CATCH_CONFIG_FAST_COMPILE` machinery, where we assume that it can
only be invoked because the assertion macros are not configured to
catch assertions.

Closes #1292

All the previous refactoring to make the assertion fast paths
smaller and faster also allows us to implement the fast paths
just with thread-local and atomic variables, without full mutexes.

However, the performance overhead of thread-safe assertions is
still significant for single threaded usage:

|  slowdown |  Debug  | Release |
|-----------|--------:|--------:|
| fast path |   1.04x |   1.43x |
| slow path |   1.16x |   1.22x |

Thus, we don't make the assertions thread-safe by default, and instead
provide a build-time configuration option that the users can set to get
thread-safe assertions.

This commit is functional, but it still needs some follow-up work:
 * We do not need full seq_cst increments for the atomic counters,
   and using weaker ones can be faster.
 * We brute-force updating the reporter-friendly totals from internal
   atomic counters by doing it everywhere. We should properly trace
   where this is needed instead.
 * Message macros (`INFO`, `UNSCOPED_INFO`, `CAPTURE`, etc) are not
   made thread safe in this commit, but they can be made thread safe
   in the future, by building on top of this work.
 * Add more tests, including with thread-sanitizer, and compiled
   examples to the repository. Right now, these changes have been
   compiled with tsan manually, but these tests are not added to CI.

Closes #2948