Added support for detecting if the normal_complex branch is in use

[lcrippa]

No description provided.

[lcrippa]

Hey @krivenko , one question since you are already in the process of remaking the pomerol data. Alexander told me he couldn't have imaginary parts of the delta function on the real axis, which by the way the coded works was understandable. I have added the possibility of adding a bit of broadening, but if I do so the tests fail.

By any chance, do you check delta on the real axis, and if yo do you set im(zeta) to zero?
If you do so, could you try with some broadening and see if it works? The relevant commit is 2169229

[krivenko]

By any chance, do you check delta on the real axis, and if yo do you set im(zeta) to zero?

In all my tests the broadening is finite. EDIpackSolver extracts GF/bare GF/Sigma/Delta data from the solver using the following code:

    def _make_gf(self, ed_func, real_freq, anomalous) -> BlockGf:
        if anomalous:
            if ed.get_ed_mode() != int(EDMode.SUPERC):
                raise RuntimeError("anomalous = True is only supported for "
                                   "superconducting bath")


        if real_freq:
            mesh = MeshReFreq(window=(ed.wini, ed.wfin), n_w=ed.Lreal)
            z_vals = np.asarray([complex(z) + ed.eps * 1j for z in mesh])
        else:
            mesh = MeshImFreq(beta=ed.beta, S="Fermion", n_iw=ed.Lmats)
            z_vals = np.asarray([complex(z) for z in mesh])


        with chdircontext(self.wdname):
            data = ed_func(z_vals, typ='a' if anomalous else 'n')

https://github.com/EDIpack/edipack2triqs/blob/c0473ec91170c11cc082e72cd85a7cad8deec14d/edipack2triqs/solver.py#L960-L974