Redeclaration of max_factorial value

[ckormanyos]

I found this one while snooping around with older compiler(s) and other projects. But it also influences Math/Multiprecision directly.

This minute change raises the question of exactly how should we deal with static constexpr POD members of a class/struct? Even in the presence of BOOST_NO_INCLASS_MEMBER_INITIALIZATION do these really need out-of-class initialization for any of the compilers we are using?

That there is a mismatch between const and constexpr is truly wrong. But what is right or less wrong?

I addressed this question in my own developments about two years ago while modernizing and came to the conclusion that the extra out-of-class initialization is redundant, harmless and somehow is handled properly (or simply disregarded) by every old, new, embedded, etc. compiler that I use from 11, 14, 17, to 20 and beyond. So I synchronized the word constexpr and retained all such initializations in my projects.

So within the context of this minute change in this PR, maybe some dialog regarding our philosophy in Boost will ensue?

Cc: @jzmaddock and @mborland and @NAThompson.

[ckormanyos]

This few-letter change is ready for review/discussion

[jzmaddock]

Looks fine to me.

[ckormanyos]

Looks fine to me.

Thank you , John!

See also Multiprecison 429