closes gh-4896

Fallback (similar to python3.4 math.log2) provides int(log(2**i)) == i
in default rounding mode.
Non-default rounding modes not tested but if it does not work we are at
least not worse than python.
Closes gh-4787

Signed-off-by: William Grant <wgrant@ubuntu.com>
Signed-off-by: Michel Normand <normand@linux.vnet.ibm.com>

closes gh-4225

ENH: avoid expensive clears in fenv functions

Clearing is 50-100 times more expensive than checking on x86, so check if there
is anything needs to be cleared first. This speeds up scalar operations
by 10%-20%.
Based on Arink Verma code in #3739.

Implement the functions as new C-API functions npy_get_floatstatus and
npy_clear_floatstatus in npy_math.

On any platform where we used our own implementation of hypot, (e.g.
windows) there were several issues with infs.

hypot(inf, y) == inf, for all y including nan.  We now check for inf and
handle this case explicitly for all y instead of just y = nan.

Although hypot(inf, 0) was correctly returning inf, it was incorrectly
raising the invalid floating point exception.

Fixes gh-2385

Now is as good a time as any with open PR's at a low.