Modifications to triangular element substitution

[SvenSerneels]

Dear Maxime

I have fixed a set of bugs in PyCopula:

• The way the upper and lower triangular elements of R were filled up , led to crashes for dim >2. I now use the much more efficient numpy routine.
• I have changed a few inconsistent calls in simulate: e.g. the Gaussian and Student copulae were called with <copula>.getCovariance(), which has to be replaced by <copula>.get_corr().
• I have fixed the indentation in Spyder to all-spaces since the original version led to inconsistencies on my machine.

Best regards
Sven