Switch the new cellarea and reproject to Proj

[asinghvi17]

This should ideally make things faster.

Benchmark MWE (on #768 with ArchGDAL):

# setup
using Rasters, ArchGDAL, Chairmarks
using Rasters.Lookups
dX = 0.0006
dY = -0.0003
lon = X(Projected(166.0:dX:167.0; sampling=Intervals(Start()), order=ForwardOrdered(), span=Regular(dX), crs=EPSG(4326)))
lat = Y(Projected(-78.0:dY:-79.0; sampling=Intervals(Start()), order=ForwardOrdered(), span=Regular(dY), crs=EPSG(4326)))
ras = Raster(rand(lon, lat))
ras2 = Rasters.reproject(ras; crs = Rasters.EPSG(3857))

# benchmarks
julia> @be Rasters.cellarea(Planar(), ras2) seconds=3
Benchmark: 634 samples with 1 evaluation
 min    3.289 ms (18 allocs: 42.518 MiB)
 median 3.646 ms (18 allocs: 42.518 MiB)
 mean   4.695 ms (18 allocs: 42.518 MiB, 7.53% gc time)
 max    185.097 ms (18 allocs: 42.518 MiB, 96.16% gc time)

julia> @be Rasters.cellarea(Spherical(), ras2) seconds=30
Benchmark: 3 samples with 1 evaluation
        12.187 s (161175590 allocs: 6.004 GiB, 5.18% gc time)
        12.258 s (161175590 allocs: 6.004 GiB, 5.32% gc time)
        12.640 s (161175590 allocs: 6.004 GiB, 7.09% gc time)
julia> @be Rasters.reproject($ras; crs = Rasters.EPSG(3857))
Benchmark: 23 samples with 1 evaluation
 min    4.075 ms (35167 allocs: 1.254 MiB)
 median 4.168 ms (35167 allocs: 1.254 MiB)
 mean   4.186 ms (35167 allocs: 1.254 MiB)
 max    4.362 ms (35167 allocs: 1.254 MiB)

With Proj:

julia> @be Rasters.cellarea(Spherical(), ras2) seconds=30
Benchmark: 8 samples with 1 evaluation
 min    3.802 s (22231194 allocs: 1.532 GiB, 3.35% gc time)
 median 3.937 s (22231194 allocs: 1.532 GiB, 3.82% gc time)
 mean   4.170 s (22231194 allocs: 1.532 GiB, 3.92% gc time)
 max    5.337 s (22231194 allocs: 1.532 GiB, 4.70% gc time)

julia> @be Rasters.reproject($ras; crs = Rasters.EPSG(3857))
Benchmark: 120 samples with 1 evaluation
 min    629.834 μs (108 allocs: 42.859 KiB)
 median 676.875 μs (108 allocs: 42.859 KiB)
 mean   772.257 μs (108 allocs: 42.859 KiB)
 max    8.718 ms (108 allocs: 42.859 KiB)

Not bad - x6 improvement in reproject, 3x improvement in cellsize

[asinghvi17]

Implemented in tiemvanderdeure#2

[tiemvanderdeure]

I think we can close this now.