Sparsestack

Vastly faster join/merge implementation (based on Numba) which makes the addition of data to `StackedSparseArray` much faster.

What's Changed
* add to_dict() method (https://github.com/florian-huber/sparsestack/pull/9)
* empty StackedSparseArray does not return .shape anymore

**Full Changelog**: https://github.com/florian-huber/sparsestack/compare/0.2.1...0.3.0

- minor naming changes to work with matchms.

- now includes all four join modes ("left", "right", "outer", "inner")
- can handle data addition as numpy arrays (now including structured numpy arrays)
- additional test cases

- Added join_type option for `add_dense_matrix()` method as well.
- Some minor fixes and linting

- fix join modes ("left", "inner", "outer"). "right" not yet implemented
- bug fixes (for join modes)
- add more tests
- sort joined sparse data by row