[hist] raise error of badly sorted edges as with variable binwidth, and check for inf/nan

[ferdymercury]

This is related to #20174 and the discussion in #20176

Should we completely disallow infinite bin edges, since it violates a bit the assumptions of under and overflow bins? (As well as complicates the implementation of GetBinCenter, Low and Up Edge, etc). And the workaround is quite easy, just setting 1e30 / 1-e30 as variable bin edge.

Or should we only allow inf for variable bin edges, and not for fixed binwidth since there it does not make sense?

[github-actions]

Test Results

    22 files      22 suites   3d 17h 34m 20s ⏱️
 3 701 tests  3 649 ✅   1 💤    51 ❌
79 426 runs  78 071 ✅ 275 💤 1 080 ❌

For more details on these failures, see this check.

Results for commit 7a477d3.