Remove Hardcoded Thresholds in Repr

Description

Repr cuts off outputs when they are going to be longer than 500 characters.

While the python docs dont say it is a requirement, they do state "For many types, this function makes an attempt to return a string that would yield an object with the same value when passed to eval"

It would be nice to be able to do something similar to numpy, where we can set the threshold ourselves, so if desired repr can be used with eval to reconstruct an object. For reference, this is how np allows users to set it

np.set_printoptions(threshold=np.inf)

Steps/Code to Reproduce

Calling repr on any objects with a large number of params

Expected Results

Full output of below (without the ...)

Actual Results

Pipeline(memory=None,
     steps=[('minmaxscaler', MinMaxScaler(copy=True, feature_range=(0, 1))), ('selectfrommodel', SelectFromModel(estimator=LogisticRegression(C=100.0, class_weight=None, dual=False, fit_intercept=True,
          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
          penalty='l2', rando...n_jobs=1,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False))])

Note the ... after random

Versions

All

Start by reproducing the long estimator repr described in the issue and locate the repr implementation and its existing tests. Determine how a user-configurable output threshold should interact with truncated repr strings. Done means the threshold can be controlled without breaking normal repr behavior, with coverage for both truncated and full output.