useTableState was an early and hasty abstraction that hasn't proved useful in many ways. Anything
you could do with useTableState, you could easily do using the same options (assuming they exist) in
the useTable hook. For this reason, state is now a first class citizen of the useTable hook, along
with more sane properties and option locations for anything pertaining to state.
* fix(use-sort-by): sorting now ignores column ids that no longer exist
* fix(use-filters): filtering no longer fails when column doesn't exist
* fix(use-sortby): filtering out invalid sortBys before sorting
The renderer function for headers, columns, cells, aggregates, filters, etc used to mix properties
from all of those contexts, including rows. Now thow contexts are located on their own reserved
properties, eg. `Cell: ({ cell: { value}, row, column, ...instance }) => value`
BREAKING CHANGE: The renderer function for headers, columns, cells, aggregates, filters, etc used
Since useColumns was relying on groupBy logic, this was code smell. I wanted useGroupBy to be able to add that logic all by itself and not have to have dependencies in the core of the table.
To fix that, I've moved the core column and row logic to the useTable hook and added a new hook 'columnsBeforeHeaderGroups' to allow useGroupBy to do what i needs in a more pure way.
