Fix race condition related to the measurement of PdfView

[j-piasecki]

It seems like in some cases, Yoga (I think) will measure the view only along one axis first, resulting in onSizeChanged being called with either w or h set to zero. This in turn starts the rendering of the pdf under the hood with one dimension being set to zero and the follow-up call to onSizeChanged with the correct dimensions doesn't have any effect on the already started process.
The offending class is DecodingAsyncTask, which tries to get width and height of the PdfView in the constructor, and is created as soon as the measurement is complete (relevant lines here and here), which in some cases may be incomplete as described above.
By delaying calling super.onSizeChanged until the size in both dimensions is correct we are able to prevent this from happening.

[WoLewicki]

I'd move it up to the rest of private variables, also, can't we use oldw and oldh from the method?

[j-piasecki]

Sure, I can move them.

Since we're not calling super method every time, using oldw and oldh could result in inconsistencies: i.e.:

• onSizeChanged gets called with w: 1080, h: 0, oldw: 0, oldh: 0
• since h === 0, super.onSizeChanged doesn't get called
• onSizeChanged gets called with w: 1080, h: 1600, oldw: 1080, oldh: 0
• super.onSizeChanged gets called - if we were to use values from the arguments we would call it with oldw: 1080 where it wasn't called with w: 1080

I haven't looked that deep to see whether it's actually used anywhere, but since those params are there I guess they are used somewhere.