refs #60: first cut of POI rendering

[incanus]

Basic POI rendering with existing style (i.e. only alcohol-shop).

• add basic marker style to stylesheet
• pass through marker style to PBF conversion
• marker/point tile parsing, bucket, buffer, shaders, etc.
• new sprite hosting for 1x/2x variants
• sprite load accepts screen scale argument

Also, opacity is supported in the stylesheet much like on WebGL but the current stylesheet doesn't use it.

I am wiiiiide open to review here as this is the most C++ code I've ever written. Outstanding issues/questions:

• Sprite is hosted on Dropbox for now since there isn't a 2x variant on the old URL. I was able to find a pre-made 2x variant in the WebGL project.
• Not sure about ramifications of actually changing transform.pixelRatio just yet. This is used in bucket.drawPoints to use the right sprite and seems like a good place to persist it. Why was it always 1.0 before?
• Sprites aren't drawing at all on OS X yet. Can't get debug variants of shaders to work, either, so it's probably something in the draw call. Sprite image properties are coming through ok.
• Went with glDrawElements instead of glDrawArrays but not sure if it's necessary. As I understand it, we do this with lines and fill outlines (and in general) in order to avoid duplication of vertices in draw calls. This probably isn't an issue with points.
• Occasional odd clipping issues.

[kkaefer]

This is sort of intentional. If fb_width is not different from width on a retina screen, there is a bug elsewhere.

[kkaefer]

Looks very good! I'll have a look and see why the points aren't rendering on OS X.

[incanus]

• On the pixel ratio stuff, IIRC I still saw 1.0 on retina devices, i.e. fb_width = width. I will double-check this, but this was part of my confusion.

• Went with glDrawElements instead of glDrawArrays but not sure if it's necessary.

  This is ok? Can you elaborate @kkaefer so I can understand a bit more?

• Occasional odd clipping issues.

  Did you see this at all @kkaefer?

• Awesome re: OS X. I will take a look too. I just wanted to get the PR in before going offline for the weekend.

[kkaefer]

It looks like there are bogus points at 0/0, I'm currently investigating on why that is the case, because they don't seem to be added to the buffer...

[incanus]

Cool re: 504b60b and glDrawArrays. Figured that might be more efficient.

Also, the degenerate shader change makes sense.

On 9710fbe, this removes the ability to colorize the marker imagery, right?

[kkaefer]

@incanus this is not yet ready as it still doesn't work on OS X, just on iOS. There seems to be a change in how gl_PointCoord works on OS X.

[incanus]

Digging into this, but does this have something to do with differences in GL_POINT_SPRITE_COORD_ORIGIN?

[incanus]

@kkaefer on OS X you need to glEnable(GL_POINT_SPRITE); first.

I'm not sure where you were with your debugging but I got past that last blocker.

[kkaefer]

Yeah, that's what I was looking for!

[kkaefer]

I think this is good to go!

[incanus]

Badassery. Nice work @kkaefer and thanks for rounding this out.