Fixed bug in quda_interface.c

[kottnad]

If-statement had no effect due to setting first_call = 1 in first branch, i.e. freeCloverQuda() was never called.

[kostrzewa]

cscs-ci run default

[kostrzewa]

As discussed offline, this should be inconsequential because of the logic within QUDA and because:

1. our clover term is independent of tm_rho (our Hasenbusch mass precon mass)
2. Mu2() (mu squared) changes as we move between the ND heavy doublet and the various monomials of the degenerate light doublet

Still, it probably is safer to use our own logic for invalidating the clover field because we are then independent of whatever might happen on the QUDA side.