Various fixes

• No more dependency on LAPACK
• partialschur also returns a list of the eigenvalues
• partialschur now respects nev: only a conjugate pair can make the method return nev+1 eigenvalues
• Better targeting: unwanted converged eigenvalues are less likely to be return by partialschur
• The method now uses Krylov-Schur rather than Lehoucq & Sorensen's implicit restart
• Some more docs.

Small API improvements and convergence detection

• The stopping criterion for near zero eigenvalues has been fixed; it was too strict.
• Errors are thrown when the input matrix is non-square or the kwargs are incorrect.
• Default parameters make more sense for small matrices.
• partial_schur is renamed to partialschur
• schur_to_eigen is renamed to partialeigen
• min and max are now called mindim and maxdim respectively.
• partialschur works with integer matrices now.

Initial release

The package is still a work in progress, but with basic functionality it might be worth it to have an official initial release 🌅.