[ENH] Bollinger bands

[fkiraly]

I noticed sktime has some tools for smoothing, but not the popular Bollinger bands used in trading.

While one can be of different opinion on the estimator, I think it is popular enough to have it in the "smoothing" section.

It should be a good first issue, given the simple nature of the estimator it should be straightforward to implement from scratch, using pandas or numpy directly.
Some thoughts:

• the q of the moving average should be a parameter
• the remember_data tag should be true, since stream use will require to remember previous data for computing the bands. So, transform should try to retrieve remembered data, and subset it to the range required to comput bands for indices in the input.

[ishanpai]

Picking up this issue as my my first issue.