TiffSaver: add overwriteIFDValue signature that takes an IFD offset

[melissalinkert]

Backported from a private PR.

This is helpful for modifying thumbnail and sub-IFDs, which aren't
included in the list of offsets returned by TiffParser.getIFDOffsets().

I wouldn't expect this to have any impact on memo files or existing tests, but it arguably still requires a minor release since it's an API addition.

[sbesson]

With this refactoring, does that mean any call to the current overwriteIFDValue(RandomAccessInputStream, int, int, Object) API will now perform a redundant check on the first IFD?

[melissalinkert]

The header will be checked twice, yes. Could refactor further to add a overwriteIFDValue(RandomAccessInputStream, long, int, Object, boolean skipHeaderCheck)? Otherwise not sure of a good and safe way to skip that check.

[sbesson]

Since we're introducing new API anyways, I think introducing

public void overwriteIFDValue(RandomAccessInputStream raf, long ifdOffset, int tag, Object value, boolean skipskipHeaderCheck)

is acceptable. From there both public void overwriteIFDValue(RandomAccessInputStream raf, long ifdOffset, int tag, Object value) and public void overwriteIFDValue(RandomAccessInputStream raf, int ifd, int tag, Object value) can delegate to this API by passing the appropriate boolean value.

[melissalinkert]

Added in b43e748

[dgault]

With the latest update to add skipHeaderCheck the API updates here look to be good.

The existing unit tests, repo tests and workflows are unaffected by the changes. I reran some local tests using the PyramidOMETiffWriter and the generated images and metadata all looked correct and no regressions found from these additions.