Sparse-dot-topn-for-blocks

- Adding the possibility to use smaller data type float32 instead of the default float64
- Adding unit tests to cover both data types
- Adding unit tests to cover return_best_ntop==False

- defragmented memory used during computation [PR-53](https://github.com/ing-bank/sparse_dot_topn/pull/53)
- Enable Github Action for unit test
- add license metadata [PR-47](https://github.com/ing-bank/sparse_dot_topn/pull/47)

- added unit tests.
- awesome_cossim_topn checks for zero input matrices.
- fix flake8 warnings.

- quick fix for nt installation problem [ISSUE-26]

- add n_jobs value validation
- matrix dimension validation
- change windows mingw32 compiler
- fix numpy and Cython install dependency

- import parallel implementation