Avoid appending redundant files to the delta table

[rtyler]

If oxbow is re-triggered with a file that has already been added to a Delta Table, it will happily add another add action to the transaction log for a redundant file.

While not a problem per se, it is unnecessary and inefficient. append_to_table should probably look at table.get_files before adding the files passed into it.

[rtyler]

Now that I'm thinking about this problem more, this might be a critical dependency to allowing oxbow to work with tables which have merges or optimizes where some files are being removed while others are being added 🤔

[rtyler]

In Lambda this can cause some duplicate data to be added to the table when fed by SQS since its delivery is not "exactly once" guaranteed, e.g.:

>>> files = dt.files()
>>> len(files)
4279
>>> len(list(set(files)))
2508
>>>