fix: Header-only fix is_valid for empty Indexed or IndexedOption

[zonca]

Noticed that the implementation of is_valid in #3901
did not handle correctly empty objects.
In that case max_index_ has the initial value of 0
the length of content is also 0.
Therefore is_valid fails because they are equal.

Added 2 test cases for this to the same file given they are related.

• test: test empty IndexedOption builders are valid
• test: suppress printout to avoid thinking tests fail
• fix: append_index() should call append_index(i)
• fix: specific case of is_valid to catch empty objects
• test: separate test case for empty Indexed and IndexedOption

[jpivarski]

This is a good fix—thanks for catching it!

I'll merge this PR as soon as we get the MacOS bug fixed. I think we'll get that done tomorrow, even if we have to put an upper limit on the pyarrow version we accept.

[ManasviGoyal]

@jpivarski Since the MacOS bug is fixed, this can be merged, right?
@zonca Once there is an awkward release with this fix, we should bump the minimum version of awkward in kaitai_struct_awkward_runtime and make a release (or maybe once you are done with the documentation).

[jpivarski]

Yes, the MacOS bug was the only thing holding it back. I'll merge it now.