Jai-sdk

- Fixed error when frequency_seconds=0 on setup method 267
- Changed split function to always return lists instead of mixed of lists and int

- [Deprecation] Usage of data as list or np.array has been deprecated, list of ids on similarity is still accepted.
- [Pipe] Series names are used on setup of Text, TextEdit, FastText and Image. The same name should be used on inference (similarity, add data and predict) making it possible even when the collection is used as a preprocessed base.
- Added jai.preprocessing.split function to split columns from DataFrames and do the id mapping.
- Changed requirements

• New describe method
• New BaseJai class to handle api requests directly
• Added support of "JAI_SECRET" environment variable with authentication key, name can be changed with var_env parameter.
• Fixed data type datetime transforming to int when converting to json

- Fix error on processing errors.

- Documentation enhance
- Fix predict processing (as_frame param on predict method)

- requests json parameter
- Changed the requests parameter to send the request's body from `data` to `json` as it will raise a error after the new API update in June 22nd for any version `jai-sdk < v0.13.0`
- Alternatively, the user can add to `jai.header` the `content-type` header as `application/json`.
- added testing in python 3.9
- addded `as_frame` parameter to predict method.
- documentation enhances