Possible bug in geotiff writing

[vpipkt]

This needs a unit test. assigning to me until write unit test.

I think I have found a bug:

• read small raster,
  • noting its dimensions
  • is in epsg:4326 or otherwise cell resolution is possibly irrational number
• write with geotiff writer using original crs and original dimensions

Expected result: new file with identical metadata to source file, as far as crs, origin, resolution, and total extent.

Observed result, small error in resolution in one dimension results in a small error in the total extent.