Add test for empty titles in DPR processor

[tholor]

When using DPR processor with use_fast=True and embed_title=True, we currently get an error if the "title" is an empty string. This is due to the handling of empty strings by FastTokenizers (was already resolved by huggingface/tokenizers#389).
Adding a test case here to cover this in future.

11/30/2020 06:53:05 - ERROR - farm.data_handler.processor -   Error message: Truncation error: Specified max length is too low to respect the various constraints
11/30/2020 06:53:06 - ERROR - farm.data_handler.processor -   Could not create sample(s) from this dict: 
 {'passages': [{'title': '', 'text': '"letters). Some ambiguity exists in the official code page assignment for the Kamenický encoding: Some dot matrix printers of the NEC Pinwriter series, namely the ""P3200""/""P3300"" (""P20""/""P30""), ""P6200""/""P6300"" (""P60""/""P70""), ""P9300"" (""P90""), ""P7200""/""P7300"" (""P62""/""P72""), ""P22Q""/""P32Q"", ""P3800""/""P3900"" (""P42Q""/""P52Q""), ""P1200""/""P1300"" (""P2Q""/""P3Q""), ""P2000"" (""P2X"") and ""P8000"" (""P72X""), supported the installation of optional font EPROMs. The optional ROM #2 ""East Europe"" included this encoding, invokable via escape sequence codice_3 with (n) = 23. While named ""Kamenický"" in the documentation, it was originally advertised by NEC as code page 867 (CP867) or ""Czech"". (However, it was never registered with IBM under that ID, as IBM registered"', 'label': 'positive', 'external_id': 'b9d68e52-72ed-414f-baed-99275dba4767'}]}