Add strlcat variant that uses memccpy for writing SDPs

[lminiero]

We recently found out that the process of creating/writing SDPs could be slow, due to how g_strlcat works. This is an attempt to speed that part up by adding a janus_strlcat variant that uses memccpy instead of g_strlcat. It seems to be working as expected in some tests when using SDP fuzzing with a huge (~20k mlines) SDP, as it's basically 10x faster:

[lminiero@lminiero janus-gateway]$ time ./fuzzers/out/sdp_fuzzer ~/Downloads/giganto.sdp 
Running: /home/lminiero/Downloads/giganto.sdp
Done:    /home/lminiero/Downloads/giganto.sdp: (410058 bytes)

real	0m3.556s
user	0m3.523s
sys	0m0.022s

[lminiero@lminiero memccpy]$ time ./fuzzers/out/sdp_fuzzer ~/Downloads/giganto.sdp 
Running: /home/lminiero/Downloads/giganto.sdp
Done:    /home/lminiero/Downloads/giganto.sdp: (410058 bytes)

real	0m0.340s
user	0m0.324s
sys	0m0.016s

and a simple test with the EchoTest demo worked flawlessly too, but of course we need to ensure this doesn't break anything in production, hence why this is still a PR and not in master yet, so feedback welcome! Once this is confirmed to be ok, I'll update the multistream branch as well where this enhancement would definitely have more of an impact.

[tmatth]

nit: to avoid confusion, probably better to log "janus_strlcat_fast:...

[lminiero]

Ah that's a copy/paste typo, thanks for catching! Incidentally, since this is a LOG_ERR, I think the name of the function is actually redundant, since it will appear in the prefix anyway, e.g.:

[ERR] [utils.c:janus_strlcat_fast:288] janus_strlcat_fast: overflow

so I'll just remove it from both methods.

[lminiero]

Merging.