Run ASan and UBsan and fix issues

[o-]

Finally I found out how to run the sanitizers from clang.

This commit fixes a ton of memory leaks, UB and memory errors. The
alignment sanitizer is still broken and thus disabled for now.

[o-]

Here are some stats for your PR:

• WARNING: the longest CI job tests_debug2 took 2.73h
• real_thing suite improved by 1.06
• real_thing_annotations suite improved by 1.06
• vectorFor regressed by 0.89
• Overall benchmarks improved by 1.01

Please find your performance results at https://rir-benchmarks.prl.fit.cvut.cz/diff?job_ids[]=1366953616&job_ids[]=1361572392&selection=all

[o-]

Here are some stats for your PR:

• WARNING: the longest CI job tests_debug2 took 3.11h
• real_thing_annotations suite regressed by 0.97
• mandelbrot_ascii regressed by 0.93
• reversecomplement_naive regressed by 0.94
• fasta_3_strict regressed by 0.94
• emptyFor regressed by 0.88
• Overall benchmarks regressed by 0.99

Please find your performance results at https://rir-benchmarks.prl.fit.cvut.cz/diff?job_ids[]=1369896929&job_ids[]=1369896978&selection=all

[o-]

Here are some stats for your PR:

• WARNING: the longest CI job test_features_3 took 2.79h
• are-we-fast-r suite regressed by 0.84
• real_thing suite improved by 1.07
• real_thing_annotations suite improved by 1.03
• Storage regressed by 0.68
• Bounce_nonames regressed by 0.77
• Bounce_nonames_simple regressed by 0.83
• reversecomplement regressed by 0.91
• spectralnorm regressed by 0.95
• pidigits regressed by 0.95
• fastaredux_naive_strict regressed by 0.89
• matrixFor regressed by 0.94
• emptyFor regressed by 0.95
• Overall benchmarks regressed by 0.99

Please find your performance results at https://rir-benchmarks.prl.fit.cvut.cz/diff?job_ids[]=1370752995&job_ids[]=1370735306&selection=all