Fix TableReader for tables without rows

[bogdankostic]

Previously, when passing tables to the TableReader that do not contain any rows, the TableReader was running into IndexError: index -1 is out of bounds for axis 0 with size 0. This PR fixes this bug by skipping tables that do not contain any rows.

CC @Timoeller

[tstadel]

LGTM! If you want you can place the check behind the declaration of the table variable. That might make the call to shape better understandable.
Besides that I trust you that a warning is the right "tool" in this situation. But I cannot really assess that. Just because I'm curious about that: In which situations does an empty table / dataframe occur? Is this a common case or is there something wrong with the "extraction" of the table?
In any case a small test to assert this behavior would be beneficial and could even help to understand in which situations the empty dataframes occur.

[bogdankostic]

I added tests and will merge this PR once the CI has passed. We have seen a case where the ParsrConverter detects only the column header row of a table inside a PDF, resulting in a table that does not contain any rows.