Rename RSHIFT macro

[ferdnyc]

Because the FFMpeg developers don't want to break their API by removing their RSHIFT definition, and it conflicts with Ruby's definition in their own header files (when building the Ruby bindings), this change jumps through a few hoops to work around the issue locally.

In a two-pronged attack, it:

• Renames RSHIFT to FF_RSHIFT in FFmpegUtilities.h, after loading the appropriate FFMpeg header files where it's defined.
• Changes the ruby openshot.i SWIG file to:
  • Temporarily rename RSHIFT to RB_RSHIFT after it's defined by the Ruby headers
  • Make sure RSHIFT isn't defined after loading the FFMpeg headers, or it renames it to FF_RSHIFT if it was
  • Restore the definition of RSHIFT to Ruby's version, from the RB_RSHIFT macro where it was earlier stored

So, after these changes:

1. FF_RSHIFT will be defined everywhere in the libopenshot codebase
2. In the Ruby bindings, RSHIFT and RB_RSHIFT will have the same definition, matching Ruby's RSHIFT
3. RSHIFT will not be defined outside of the Ruby bindings

We never actually use either RSHIFT anywhere in the code, so this shouldn't affect anything — it's purely to silence compiler warnings (and ensure that we don't accidentally end up using the wrong RSHIFT in some of our code). The pieces of FFMpeg's libavcodec that do use their RSHIFT (there are a few codecs) have already been compiled into the library, so they don't rely on the macro definition anymore and won't care that we've #undef'd it.

If we ever want to use FFMpeg's rounding RSHIFT, we just have to remember to call it FF_RSHIFT.

Fixes #164

[ferdnyc]

@jonoomph I think we should still do this, as well.

Like the Point type conflict you hit on macOS — which you fixed basically the same way — the conflict between FFMpeg's RSHIFT and Ruby's RSHIFT isn't likely to go away anytime soon. (Not after I struck out getting the FFMpeg people to rename their macro.)

So, while in a sense this may just be a matter of eliminating paranoid compiler warnings, really it's protection against anyone accidentally using the wrong macro somewhere down the line, since with this change RSHIFT() won't be defined in our headers, and anyone wanting to use FFMpeg's version of it will have to explicitly call FF_RSHIFT().

(I suppose if we want to be really paranoid, we could define RSHIFT() to spit out a compiler warning to anyone who attempts, instead of just leaving it undefined.)

[jonoomph]

LGTM! Merging (I'm a little worried this might break something on our builders, so let's see what happens). 😄