Support error type for formula cell

[EnverOsmanov]

When a cell has formula inside - LibreOffice shows it empty, but spark-excel silently skips(!) these rows.

@waiyan1612 could you explain, please, why PlainNumberReadSuite expects to skip rows? See failed test cases.

[waiyan1612]

Hi @EnverOsmanov, when I was developing the test suite, I just followed the existing behaviour of the inferSchema. That's why in the suite, using inferSchema=true is expecting 4 rows but inferSchema=false is expecting 5 rows.

Perhaps @nightscape can provide a better context. We will also need to consider the backward compatibility of this change.

[nightscape]

The "correct" solution will land with #389 where one can specify a ParseMode which determines how to handle rows with corrupt data.
@quanghgx is actively working on that branch.
I'd be ok with merging this as a temporary solution after the tests are adapted accordingly.

[EnverOsmanov]

@waiyan1612 , thanks for update!

The test was adjusted to not skip rows.

Regarding backward compatibility - I believe skipping rows is a bug and shouldn't be an expected behavior. So it should be fine "break" such behavior.

[nightscape]

@EnverOsmanov LGTM! Feel free to squash merge 👍