[tables] Update extract simple table with option to allow gaps

[jstockwin]

This is because extract_simple_table is much faster than extract_table,
and we often have cases where only a few columns or rows don't have data
in every cell.

Initially we wanted this to only be used on full table, since it can
easily miss cells. We still achieve this when allow_gaps=False
(default), but you can say that you're expecting gaps.

Closes #58

[jean-garret]

I have a few questions (see comments) and an oversight. Could you also test the behaviour of extract_simple_table when passing a reference_element.

[jean-garret]

Should we test that reference_element is actually in elements (if it is not None) or do you think it would be unnecessary?

[jstockwin]

Hm, interesting point.

Since writing this I've realised that I can still check that the number of elements passed equals the number of elements in the returned table (in commit "[tables] Fix bugs in extract simple table"):

    table_size = sum(
        len([element for element in row if element is not None]) for row in table
    )
    if table_size != len(elements):
        raise TableExtractionError(
            f"Number of elements in table ({table_size}) does not match number of "
            f"elements passed ({len(elements)}). Perhaps try extract_table instead of "
            "extract_simple_table, or change you reference element."
        )

I think you can still get away with passing a reference element which isn't in the list, but perhaps if you do this it's intended. I think it's fine not to check, and then if for some reason you don't want your reference element included in the final result then that's fine? I think it would be hard to do without noticing (i.e. if someone does it then it is probably intentional).

What do you think?

[jean-garret]

I agree. I think it is fine not to check.

[jstockwin]

@jean-garret I've updated this PR

[jean-garret]

LGTM. Nice that you spotted the bug with extract_simple_table when the table is across different pages!

[jean-garret]

I agree. I think it is fine not to check.