`typeof(missingval)` should be a subtype of `eltype`

[mauro3]

Consider

ras = Raster(rand(Int32, X(25:1:30), Y(25:1:30)), missingval=-1.5)

Is allowing the type of missingval not being a subtype of the eltype useful for anything?

The reason I came across this is that it took me ages to figure out what was wrong here:

julia> ras = Raster(rand(Int32, X(25:1:30), Y(25:1:30)), missingval=-1)

6×6 Raster{Int32,2} with dimensions: 
  X Sampled{Int64} 25:1:30 ForwardOrdered Regular Points,
  Y Sampled{Int64} 25:1:30 ForwardOrdered Regular Points
extent: Extent(X = (25, 30), Y = (25, 30))
missingval: -1
parent:
              25           26           27           28          29           30
 25  -1577965140   -930532794  -1069034332  -1869000944  -876112444  -1896541373
 26  -2119450237   2112008256   -574130295    234237348   295379253  -1608673699
 27    479548452   1163904508  -1753155508   -779730610  -508137957   -545965278
 28   -518000084  -1234244315   2027127016  -1383270083  1182867062   1198615452
 29   -575807158    253215731    731658899  -1561393151  2005117126   1945068211
 30  -1836660204  -1824357014   -734002877    853304731  1129518778  -1924398571

julia> write("/tmp/band.tif", ras, force=true)
ERROR: GDALError (CE_Failure, code 1):
        GetNoDataValue() should be called instead

Stacktrace:
...

which can be fixed with

julia> ras = Raster(rand(Int32, X(25:1:30), Y(25:1:30)), missingval=Int32(-1));
julia> write("/tmp/band.tif", ras, force=true)

Anyways, should there be a conversion missingval = convert(eltyp(data), missingval)?

[rafaqz]

Its probably not useful for anything.

And yes that could trigger the wrong gdal methods (it has specific methods for different types, which is kinda crazy)

Lets fix it to match the array eltype

[rafaqz]

This is enforced now