write.nrrd should be able to use header for core fields as well

[jefferis]

Those specified as main arguments should still take priority. Should also normalise argument/field names

e.g.

enc-> encoding
dtype->type 

though for dtype, should probably keep a deprecated alias.

[jefferis]

May also need to change dtype in write.amiramesh. Or accept that we have dtype which accepts a subset of type names and use header$type to set using any nrrd type name.

[jefferis]

I think that we could have different options for dtype in write.amiramesh and write.nrrd. But the option of setting fancier type values in header$type is also fine I think.

[jefferis]

One issue to think about carefully if this is implemented is handling of default type, endian and encoding values which are carried forwards from nrrds that have been read in with read.im3d. For example if they have been modified the historical default of writing as dtype="float" might be more appropriate than using the original type.