Hhpy

hotfix for package info
security update bump bleach

- hhpy.ds
- k_split now supports splitting timestamps by a pre-defined value

it's been a while so I forgot how to do releases... bugfix release

- hhpy.modelling
- can now specify multi output postfixes
- bugfixes

- hhpy.modelling
- class Model not supports setattr, a function that sets an attribute of all k instances of it's model
- support for multi output one input regressors with attribute multi
- various bugfixes
- hhpy.ds
- function top_n supports passing a percentage with a max number
- convenience function reformat_columns to unify column names
- reworked df_score to work with the y_true, y_pred format

- hhpy.regression (new module)
- Conv1DNN model: wrapper for keras.Conv1D
- hhpy.main
- new functions
- copy_function: returns a copy of a function (useful for defining repr rules)
- changes:
- renamed 'force_list' to 'assert_list' for consistency
- rename 'force_scalar' to 'assert_scalar' for consistency
- bug fixes:
- fixed repr
- hhpy.ds
- new functions
- rolling_lfit: calculate the rolling linear fit over a pandas Series
- rank: rank a DataFrame column / Series according to criteria, supports groupby
- hhpy.plotting
- new functions
- cat_to_color: Encodes a categorical column as colors of a specified palette
- hhpy.modelling
- k_split, train, predict and fit now support fit_type 'k_cross'
- automatically split the data into k sub splits and cross validate
- predict can choose between k_predict_type 'test' (return test prediction for each k group) and 'all' returns
prediction for each k group
- new functions
- to_keras_3d: Turns a pandas DataFrame into a 3D array required by Keras