Exclude signal handler from linux crashlogs

[NicksWorld]

Currently all crashlogs produced on Linux end with a opaque offset within libdfhack.so, which can cause some confusion.

Signal SIGSEGV
0> hack/libdfhack.so(+0x936f90) [0x7b040d336f90]

This PR causes the crashhandler to skip printing the first backtrace entry, which will always be dfhack_crashlog_handle_signal or dfhack_crashlog_handle_terminate, where the backtrace is retrieved. To ensure exactly one crashhandler-related entry is added to the stack frames, handle_signal_internal is marked with [[gnu::always_inline]] to ensure it doesn't create a stack frame.

Alternatively, we can clear up the opaque entry by enabling symbol visibility on dfhack_crashlog_handle_(signal|terminate) by marking them with DFHACK_EXPORT, with the following result:

Signal SIGSEGV
0> hack/libdfhack.so(dfhack_crashlog_handle_signal+0x62) [0x7fffe3955052]

[ab9rf]

this code will print nothing if -- somehow -- the backtrace only has one entry. this can happen if the stack is corrupt at the time of the crash. i would advise printing a "omg disaster happened" message in the event that crash_info.backtrace_entries has fewer than two entries, rather than printing nothing as this is potentially slightly less confusing

the none at all case is handled above, and probably should also emit some sort of omg message, again to reduce user confusion. maybe something like "stack trace unavailable". if the API used to acquire the stack trace has any error indicators, these should probably be emitted when they obtain