Correctness fixes and (optional) better C stdlib integration

[jlindgren90]

• Add early NULL check to ta_free()
• Add overflow checks in alloc_block() and ta_calloc()
• Clear full block size to 0's in ta_calloc()
• Optionally set errno to ENOMEM when out of memory
• Optionally use memset() to implement memclear()
• Add ta_getsize() to return allocated block size
• Add ta_realloc() (optional, requires memcpy())

ta_getsize() and ta_realloc() are based on ideas from #11, but rewritten to fix various issues in that PR.

[jlindgren90]

Updated with some more fixes:

• Fix behavior of ta_alloc(0) (see #15)
• Fix additional arithmetic overflows
• Add fallback memcopy() when not using C stdlib
• Add early return to ta_realloc() when size is similar

[postspectacular]

@jlindgren90 Thank you very much for this and the other PR - I will take a look asap, but likely will take a little while for me to review & test this...