Multi-part EnumType extraction

[duklin]

Is your feature request related to a problem? Please describe

So far, only the simple enums have been covered such as:

algorithm : {'arpack', 'randomized'}, default='randomized'
gamma : {'scale', 'auto'} or float, default='scale'

There are multi-part enums whose values are inside the docstring description such as:

loss : str, default='hinge'
The possible options are 'hinge', 'log', 'modified_huber', 'squared_hinge', 'perceptron', or a regression loss: 'squared_error', 'huber', 'epsilon_insensitive', or 'squared_epsilon_insensitive'

learning_rate : str, default='optimal'
- 'constant': `eta = eta0`
- 'optimal': `eta = 1.0 / (alpha * (t + t0))` where t0 is chosen by a heuristic proposed by Leon Bottou
- 'invscaling': `eta = eta0 / pow(t, power_t)`
- 'adaptive': eta = eta0, as long as the training keeps decreasing

strategy : str, default='mean'
If "mean", then replace missing values using the mean along each column
If "median", then replace missing values using the median along each column
If "most_frequent", then replace missing using the most frequent value along each column
If "constant", then replace missing values with fill_value

Desired solution

Expand the current logic for EnumType extraction to include the multi-part enums.

Possible alternatives (optional)

Screenshots (optional)

Additional context (optional)