Add the BMTI method for density estimation

[AldoGl]

Proposed changes

In this PR we:

• add the kstar module. This contains methods related to the computation of kstar and is now inherited by density_estimation.
• add the neigh_graph module. This contains methods related to the computation of the neighbourhood graph and it inherits from kstar. These methods are used in BMTI but are more general and thus deserve a separate class.
• add the density_advanced module. This essentially contains the BMTI method, and it inherits from density_estimation and from neigh_graph.
• add a jupyter tutorial notebook to illustrate the BMTI method.
• add tests for the new methods.

Work done with @charliematteo.

Types of changes

What types of changes does your code introduce? Put an x in the boxes that apply

• [ ] Bugfix (non-breaking change which fixes an issue)
• [x] New feature (non-breaking change which adds functionality)
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)

Checklist

Put an x in the boxes that apply.

• [x] Lint and unit tests pass locally with my changes
• [x] I have added tests that prove my fix is effective or that my feature works

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 87.29097% with 38 lines in your changes are missing coverage. Please review.

Project coverage is 81.61%. Comparing base (7a1f695) to head (c2a1a96).

Files	Patch %	Lines
dadapy/density_advanced.py	80.85%	27 Missing :warning:
dadapy/neigh_graph.py	90.43%	11 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #119 +/- ## ========================================== + Coverage 81.07% 81.61% +0.54% ========================================== Files 17 20 +3 Lines 3006 3285 +279 ========================================== + Hits 2437 2681 +244 - Misses 569 604 +35 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[AldoGl]

@charliematteo, I did what I planned. For some reason the test "test_density_BMTI()" runs perfectly on my machine, also using tox, but when it runs here it returns an error in the solution of the linear system. Very strange, it might be machine dependent. I can change the dataset tested and check that that is the problem.

[AldoGl]

Ok, I fixed that issue. Now we have the problem that python 3.7 gives different results for some routines, also a very strange one

[AldoGl]

Ok, I managed to fix that as well @charliematteo

[AldoGl]

I think this work is now ready to be merged. It would be useful if another developer had a quick look to check whether the logic of the new modules is clear. @diegodoimo you were potentially interested in the method, would you mind checking it out? (very rapidly, when you have time!)

[diegodoimo]

I'll do it in the weekend

[AldoGl]

Thank you @diegodoimo, and don't worry too much about it!

[diegodoimo]

A minor point: I do not understand why there is compute_density_BMTI which just calls compute_density_BMTI_reg hardcoding some kwargs. Could this be in a single function instead of having two? In the future, we should also consider unifying the two density classes in a single density interface. Anyway, it looks good and can be merged to me; great job!

[AldoGl]

Thank you very much @diegodoimo. The idea was to remove unnecessary arguments for the standard BMTI, but probably you are right and it's better to simply use sensible default values for unused arguments. I will implement this change. Thank you again!