DOC: complex conjugation in BLAS functions

Is your feature request related to a problem? Please describe.

I have a very large complex array A and I want to apply its adjoint (conjugate transpose) to a vector y. If I use A.conj().T @ y then A.conj() is allocated explicitly. That consumes too much memory. It's possible to bypass allocation by calling scipy.linalg.blas.zgemv(1.0, A, y, trans=2).

The trouble is: the meaning of the trans parameter isn't specified in SciPy's zgemv documentation. I only learned of this solution by guessing the behavior of trans and trying it out.

Describe the solution you'd like.

SciPy's documentation for complex BLAS functions should indicate how to perform the conjugate-transpose operation. I.e., explain the meanings of trans = 0, 1, or 2 in documentation for the relevant functions.

I think the change is reasonable because people who are familiar with common C or FORTRAN BLAS interfaces expect trans to be a character (N, T, or H).

Describe alternatives you've considered.

N/A

Additional context (e.g. screenshots, GIFs)

Code snippet showing undocumented functionality:

import numpy as np
import scipy.linalg as la
import scipy.linalg.blas as blas

A = np.random.randn(3,2) + 1j*np.random.randn(3,2)
y = np.random.randn(3) + 1j

inefficient = A.T.conj() @ y
efficient = blas.zgemv(1.0, A, y, trans=2)
print(la.norm(inefficient - efficient))