Update business logic

This update has several breaking changes, cleaning up the pymapd internals as well as requiring pandas and arrow instead of having them as optional.

• Fixes for boolean na types, typos, na assignments in test cases (#110)
• Option to chunk Pandas columnar data load (#117)
• Clean-up branding (#121)
• Remove NumPy shim, replace with pa.foreign_buffer (#111)
• Make NumPy, pandas and pyarrow hard deps (#120)
• Datetime bounds fix (#123)
• Null fixes on columnar load (#125)
• Fixes for FLOAT and DOUBLE nans (#127)
• Get types from existing schema on load (#128)
• Option to load by matching table column names to dataframe column (#129)