Improvements for env installs

• removed unused packages: dask, pillow, statsmodels, sqlalchemy
• clarified reqs w.r.t mamba install MKL

Minor improvements to sampling and plotting

• made sample stepper explicit
• minor improvements to model plotting
• minor improvements to eda describe, display

Minor tidyup

No great changes from v0.4.0 just a general tidyup

Ready for a new point release

• New model reload capability
• Improved file handling throughout
• General consolidation throughout

Minor improvements to file handling

• Minor updates to requirements
• improved curate.SimpleStringIO

Updates to BasePyMC3Model to allow model "reload"

• Overhauled BasePYMC3Model to save traces to idata only
• Overhauled _create_idata accordingly to allow creation from prior, trace or posterior idata with logical dependencies
• Overhauled update_idata accordingly to allow attaching an externally loaded idata object (model reinstantiation)
• Renamed all relevant functions to use idata rather than azid

Minor improvements to reduce package size

• dropped nltk and related text processing - will create a new package for that if needed
• general tidy of unused imports
• cleaned up inits

Added new custom distribution for ZeroInflatedLogNormal

• Added new custom distribution model.ZeroInflatedLogNormal
• Minor improvements to model plots

Stability update

removed unused packages
updated orderedset to ordered-set (the maintained replacement)
autoupdate precommit

General widespread updates to packages and minor new functions under eda

Packages

• updated package versions, primarily for pymc3 v3.11.4 and dependencies
• added hypothesis to allow pymc3 testing

Data Curation

• improved azid I/O
• improved datatype converter (incl. improved logic for bool datatype conversion)

PyMC specific

• included builder for inference data
• improved idata and write idata to file
• corrected samples default for ppc
• allow ppc be returned in an idata object or attached to the model object
• added better graphviz handling

Utilities

• moved code to embed pngs as matplotlib images (so they render on github) to new eda.utils.py
• added code to model.utils.py to call model render to graphiz, and insert dpi in the case of png and save to file
• combined these let you create and display graph representation in a Notebook on Github which is handy
• default plots to a reasonable size
• renamed display_headtail
• added datadict writer, and improved fts_by_dtype
• added new simple month_diff