Foostrap

Refactored the numba compiled functions to drop the use of functions as arguments. This allows the correct caching and reusing of the functions in separate programs (as long as the package versions and CPU are preserved), in order to save execution time for all future uses, but it makes the code a bit repetitive and less modular.

* Added more statistics:
For 1D: standard deviation, quantiles
For 2D: weighted mean, pearson correlation
* New optional arguments in the foostrap function for the statistic and 'q' for the quantile statistic
* Added corresponding unit tests for new functionality
* Optimized all the statistics with low level numba for loops
* Reorganized the code to better abstract the different statistics and the possibility of passing arguments to it

1. Fixed a bug where mixing arrays of different sparsity would give nan results
2. Added corresponding unit tests to detect nan outputs
3. Added a Jupyter notebook with comparisons against the Scipy implementation

First public release includes unit testing on minimal dependencies (Python 3.9, numpy 1.25, numba 0.58, scipy 1.11)

Features:
4 statistics available for now
Parallelism implemented
Handling of sparse and binary data implemented
BCa method for CI implemented
One-sided CIs and changing confidence level implemented