Resolve several TODOs where tag values were cast from u32 without checks
by providing generic methods that error when the value fails casting to
the desired type. In addition to being more correct, this simplifies and
shortens decoding logic.
Remove `find_tag_u32` and `find_tag_u32_vec` as they are now redundant
and a GitHub code search found no usages outside copies of
`image-tiff`.
Make the behavior of RGB and RGBA like Gray in returning ColorTypes for
all homogeneous bit depths, even if those depths are not 8, 16, 32 or
64. Note that this does not change which color types can actually be
decoded.
Note that this adds the LONG8 data type from BigTIFF because of bounds
requiring TiffValue. However, this datatype should not actually be
written to tags.
Andrew Champion