CLN/BUILD: Fix clang warnings on build

[TomAugspurger]

Here's the output from the current master. I could try to clean some of the warnings up someday if there's any interest; a decent excuse to look through the source code.

A lot of the warnings look similar:

• deprecated numpy API
• warning: static variable 'PyArray_API' is used in an inline function with external linkage [-Wstatic-in-inline]
• warning: comparison of constant -1 with expression of type 'PANDAS_DATETIMEUNIT' is always true
• warning: format specifies type 'long' but the argument has type 'long long' [-Wformat]
• various unused functions

What do things look like with gcc (and whatever windows uses)?

[cpcloud]

1. Deprecated NumPy API is on all system AFAIK
2. Not sure
3. Looks like it could potentially lead to a bug and should be fixed or at least documented.
4. Format string warning that would be nice to get rid of.
5. Unused functions/variables are probably okay to remove unless they're from Cython, which you can't really control. I would expect most compilers remove dead code, so this isn't that big of a deal.

[jtratner]

Also, klib may create unused functions because of macros...

[jtratner]

btw - you can see what the build warnings are on Travis too (just expand the build part of the build summary)

@jreback, numpy is moving on, should we put effort into moving pandas into 1.7 cleanliness?
is it delicate or just tedious?

[jreback]

I have no idea.....scipy/numpy builds are not clean themselves....does it bother anyone?

aesthetically? a little bit. there are increasingly more warnings related
to compiling pandas or running the tests. a dose of anti-entropy wouldn't
hurt, but it's not actually damaging.

[jreback]

would take someone whom is more familiar with the build process and some time

[jreback]

so we prob want to have this as an option, but just passing -w in compilation of extensions basically suppresses almost everything: https://travis-ci.org/jreback/pandas/jobs/115105866

What we could do is have an option to turn this on/off (default on). any thoughts

@TomAugspurger @jorisvandenbossche @wesm