[ENH] EnbPI and SPCI algorithms

It would be nice to have a clean implementation of the "official" EnbPI algorithms in sktime or skpro. While we do have conformal prediction algorithms, these are not the "originals".

Some of the algorithms add intervals on forecasting methods, some are tabular and then applied to time series by reduction.

List of methods:

• MultiDimSPCI - multivariate forecasting, sktime
• SPCI - forecasting, sktime
• EnbPI - this is an online tabular proba mode adder for tabular regression, with batch update, so should go in skpro

One a side note, EnbPI is indirectly interfaced via the dependency chain sktime -> skpro -> mapie, but in the mapie regressor it is part of a somewhat monolothic, complex estimator.

FYI @hamrel-cxu, are you active on GitHub? I suppose so due to recent contribs to MultiDimSPCI?

sktime follows a distributed algorithm ownership policy, so you could contribute the algorithms yourself, or review a contribution, or commit to maintenance and become an owner. Owners and authors of code are credited in the estimator overview: https://www.sktime.net/en/latest/estimator_overview.html

Another option could be turning the three mentioned repositories - EnbPI, SPCI, MultiDimSPCI - into packages with compatible interfaces, and listing the estimators in sktime. These would have to be maintained separately as packages, and give you more granular control over testing, CI, license, etc.