Add support for the . operator (and related work)

[matthewwardrop]

This is a relatively invasive and large patch set that adds support for the . operator in formulae (and potentially in the future support for other types of operator dynamism driven by execution context). Since the . operator depends on the data context, this weakens the strict separation of parsing and materialization, but this patch retains this distinction and simply requires users to pass in the context twice if using the lower level API; or does it for the user if they are interacting with formulaic using syntactic sugar methods like model_matrix. If you are not using operator dynamism, then you can completely ignore this functionality and keep the clean separation as in times of yore.

Example usage:

import pandas
from formulaic import model_matrix
df = pandas.DataFrame({"x": [1,2,3], "y": [4,5,6], "z": [7,8,9]})

model_matrix("y ~ .", df)

outputs:

.lhs:
       y
    0  4
    1  5
    2  6
.rhs:
       Intercept  z  x
    0        1.0  7  1
    1        1.0  8  2
    2        1.0  9  3

---

More complicated usage is also supported:

model_matrix("y ~ .^2 - z", df)

outputs:

.lhs:
       y
    0  4
    1  5
    2  6
.rhs:
       Intercept  x  z:x
    0        1.0  1    7
    1        1.0  2   16
    2        1.0  3   27

---

closes: #175
related: #177