Ipfx

Changed:
- Adds new 'Stimulus contains NaN values' tag and error handling to qc_feature_extractor

Bug fixes:
- Fixed memory leak in method of `EphysNWBData` caused by `lru_cache` decorator

Changed:
- Add features_state information to the pipeline output json
- Improve performance of loading sweeps from NWB2 files by using LRU cache
- More robust error checking when loading time_series

Bug fixes:
- Fix segment length rounding error in the DAT file converter

- IPFX is now [on pypi](https://pypi.org/project/IPFX/)! You can install it by typing `pip install ipfx` into your terminal.
- IPFX now supports [NeurodataWithoutBorders](https://www.nwb.org) version 2 in place of version 1.
- IPFX supports Python 3 in place of Python 2.
- Numerous features, [documentation](https://ipfx.readthedocs.io/en/latest/) updates, and bugfixes have been incorporated into IPFX.