Document BN_abs_is_word(3).

Document BN_abs_is_word(3).

While here, add the missing "const" qualifier to the second parameter
of BN_is_word(3) - even though i doubt that marking an integral type
parameter as "const" serves any significant purpose...

Note that the OpenSSL documentation for this function that Billy
Brumley committed on August 4, 2021 is actually wrong.  Essentially,
it says "BN_abs_is_word() test[s] if a equals ... |w|."  Now pray
tell me, what exactly is the point of taking the absolute value of
an unsigned integer number?  To compensate for the obvious absurdity
of his patch, Billy made a point of getting *three* OKs from rather
notable people: Pauli Dale, Nicola Tuveri, and Dmitry Belyavskiy.

I believe this is a striking example of the cavalier attitude some
projects put on display when it comes to documentation, and also a
striking example of how bad documentation can occasionally be worse
than no documentation at all, because the OpenSSL manual page will
now thoroughly confuse anyone reading it.

SCNR pointing out this (hopefully unintentional) hilarity - or is
this an attempt at trolling the readers of their documentation?
If it is, they certainly got me.