Fix segmentation faults when errors occur while handling unserialisable objects

[JustAnotherArchivist]

Errors during __repr__ itself as well as ones during the conversion to a bytes object were not handled, resulting in NULL pointer dereferencing.

Cf. #382

[codecov-commenter]

Codecov Report

Merging #531 (5a52c12) into main (7799498) will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #531      +/-   ##
==========================================
+ Coverage   91.71%   91.75%   +0.04%     
==========================================
  Files           6        6              
  Lines        1810     1819       +9     
==========================================
+ Hits         1660     1669       +9     
  Misses        150      150              

Impacted Files	Coverage Δ
python/objToJSON.c	90.43% <100.00%> (+0.06%)	⬆️
tests/test_ujson.py	99.60% <100.00%> (+<0.01%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7799498...5a52c12. Read the comment docs.

[JustAnotherArchivist]

To repeat my comment from #382: "~E~" as the error handler was, as far as I can tell, never a thing. With the "strict" error handler, this behaves much more reasonable. If __repr__ returns an invalid strict UTF-8 string (like one containing a surrogate escape), this is bubbling up now. Note that the json standard library instead raises a TypeError like 'Object of type A is not JSON serializable', but doing this from C is more involved, and I'm not sure it's the right thing to do anyway as it can hide the actual exception that caused things to blow up.