Try (with minimal effort) GCC4 in arithmetic tests

[ckormanyos]

No description provided.

[ckormanyos]

The result of this PR so far is:

• Do not include GCC 4.8 runs in CI.
• Make a separate file detail/check_cpp11_config.hpp capable of local specialization if needed.
• Repair numerous instances of mismatch const-versus-constexpr.
• Repair consistency in all header guards according to use BOOST_MP_-plus-filename-_HPP and optionally add uniqueness (e.g. _DETAIL) as needed.

Rational why I include the final point here is because I noticed this point with great clarity for the first time here.

[ckormanyos]

Hi John (@jzmaddock) this thing is review-ready. I'll struggle no further on it until we reach a 4.8 consensus. At the moment there is no effort to fully incorporate 4.8.

This PR in this state officially closes #429. You can, however, also choose to pursue this thread further. I'm OK with that too. Or later get back on it, or whatever?

I do, however, really want the changes within and as sketched/described above. Also a few changes in Math needed, as these are bugs anyway, I'll PR them shortly.

If you get a chance, please at your convenience, tell me what you think...? And we can move on.

Cc: @mborland and @NAThompson

[ckormanyos]

Any feedback on this? It seems to be running green in CI.

[jzmaddock]

Sorry for the delay, yep, go for it!

[ckormanyos]

yep, go for it!

Thanks John. This PR ended up to not go entirely for GCC 4.8 and there is no 48. added run in CI. Multiprecision was pretty close to accepting 4.8, but a few small issues prevented that.

GCC 4.8 did find some inconsistencies (mostly const-versus-constexpr) as mentioned above and this PR resolves these nicely.

Cc: @mborland and @NAThompson