Fix missing stacktraces when using -d:useNimRtl

[idlewan]

When using -d:useNimRtl, stack traces are useless ("No stack traceback available").
This can be verified when compiling this example program in addition to libnimrtl.so:

template fuzzy(x) =
  echo x[] != 9

var p: ptr int
fuzzy p       # will segfault on purpose, however "no stack traceback is available"

Compile with nim c -d:useNimRtl test_segfault.nim and nim c -o:libnimrtl.so nim-repo/lib/nimrtl.nim.
Start (on linux) with LD_LIBRARY_PATH=/path/to/libnimrtl.so ./test_segfault

This happens because two instances of framePtr exist at the same time: one from libnimrtl.so, and one from the test program (I ended up seeing this when printfing my way around and checking the generated c files). There is also two signalHandler present, the one from test_segfault is the one that runs last, apparently.

The right framePtr from libnimrtl.so is updated correctly by nimFrame and other functions, however when for example a segfault happens, the signalHandler from the test program runs, which refer to a nil framePtr.
That framePtr doesn't contain any information on frames, thus why no stack traceback were available.

---

This change ensure that there is only one signalHander that will be registered, the one from libnimrtl.so, this should also avoid pulling a second nil framePtr by the test program.
The stack traces should now be available and shown when using useNimRtl.

[Araq]

Wonderful analysis and bugfix!