… than passing TIFF header offsets around everywhere.

Note that while BigTIFF supports files greater than 2 GiB in size, our current implementation does not due to the pervasive use of Int32 throughout the code to represent offsets into the data.

This will only ever be a 16-bit value.

This allows combining the add and test operations into a single lookup.

If the byte order changes during extraction, a new reader is made via WithByteOrder.
That new reader is naturally scoped to the sub-reading operation, and the byte order
does not need to be reverted.

We have been steadily adding parameters to members of ITiffHandler. We are about to add more to support GeoTIFF. Rather than do that, introduce a new type that encapsulates this common state. This makes the API simpler, and means we can add members in future without it being a breaking change.