Issue 5658: Include <cmath> consistently, not <math.h>

https://sourceforge.net/p/testlilyissues/issues/5658/

In the past, we had problems with std::isinf and std::isnan not being
provided by <cmath> because they are C++11 features, so we were
including <math.h> in some places, and even potentially defining our
own isinf.

Now, we are using C++11.  Because the code is also graced with "using
namespace std;" all over the place (see issue 4550), ambiguity may
arise.  Using <cmath> consistently and removing our local definition
of isinf should improve the situation.

[Dan Eble, 2020-01-10 00:32:01]

Remove our own definition of isinf()

[lemzwerg, 2020-01-10 05:13:57]

LGTM, thanks.

[michael.kaeppler, 2020-01-10 14:05:10]

Passes 'make all' on my LilyDev1-image in VirtualBox. 
Regtests against current master (	e43878e2d40542bf492c7468fa68513162bc5972) as
baseline do not show any differences.
Now waiting for GUB build to complete.

[Dan Eble, 2020-01-11 02:54:48]

Qualify isinf, isnan, isfinite with std::

[Dan Eble, 2020-01-11 02:55:42]

Cleanup: !isinf && !isnan == isfinite