refactor(selectors): remove janky Predicate class and unify Selectors under a single interface

[cpcloud]

Description of changes

This is a refactor of selectors internals.

The disparity between the Predicate class and the Selector has always
bothered me, because not every selector can be easily made into a predicate.

And, for the ones that can, there are some often-used ones that are really
awkward to express as predicates, like the c selector, which expressed as
a predicate led to performance problems and IMO unnecessarily complex solutions
to that performance problem.

I thought to myself, "What are selectors doing fundamentally?"

And the answer I came up with is that they are producing a sequence of column
names that can be used to subset columns in a table.

Set operations ended up being a much more natural way to model the problem, while
allowing subclasses of Selector to customize operations for whatever reason
e.g., avoiding any overhead if there's a fast path (selecting no columns) or if
overriding is necessary for the functionality (e.g., And).

So, I turned Predicate into Where, and rewrote a bunch of the selectors
that were previously based on where into their own specialized classes, and
boiled everything down to expanding the names of a selector based on an input
table, and rewrote the top-level expand method in terms of name expansion.

Previously we were also incurring the overhead of table column selection in
every case, and for a good chunk of selectors, we can avoid that overhead by
computing directly on the table's columns.

[cpcloud]

Since we benchmark selectors for the wide table case, I ran them, and there's a consistent 10% improvement.

Nothing to write home about, but I'll take the win.

-------------------------------------------------------------------------------- benchmark 'test_selectors[10000]': 2 tests --------------------------------------------------------------------------------
Name (time in ms)                             Min                 Max                Mean            StdDev              Median                IQR            Outliers     OPS            Rounds  Iterations
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_selectors[10000] (0001_949fbea)     919.4184 (1.10)     937.7005 (1.11)     928.4652 (1.10)     8.3591 (2.51)     927.5156 (1.10)     15.7539 (2.96)          2;0  1.0770 (0.91)          5           1
test_selectors[10000] (NOW)              837.6641 (1.0)      845.8418 (1.0)      841.4304 (1.0)      3.3308 (1.0)      840.3802 (1.0)       5.3216 (1.0)           2;0  1.1885 (1.0)           5           1
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

------------------------------------------------------------------------------- benchmark 'test_selectors[1000]': 2 tests -------------------------------------------------------------------------------
Name (time in ms)                           Min                 Max               Mean             StdDev             Median               IQR            Outliers      OPS            Rounds  Iterations
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_selectors[1000] (0001_949fbea)     86.8199 (1.13)      87.4022 (1.0)      87.0923 (1.05)      0.1968 (1.0)      87.0289 (1.12)     0.3323 (1.0)           3;0  11.4821 (0.95)          9           1
test_selectors[1000] (NOW)              76.9421 (1.0)      110.7883 (1.27)     82.6303 (1.0)      12.3097 (62.55)    77.4166 (1.0)      1.0610 (3.19)          2;2  12.1021 (1.0)          13           1
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Legend:
  Outliers: 1 Standard Deviation from Mean; 1.5 IQR (InterQuartile Range) from 1st Quartile and 3rd Quartile.
  OPS: Operations Per Second, computed as 1 / Mean
========================================================================================= 2 passed in 9.53s ==========================================================================================

[jcrist]

Overall this looks like a nice improvement!