Investigate tolerance in chol_omp

[RichardWaiteSTFC]

When I run the incommensurate BiFeO3 calculation with mex disabled and hermit=True I get this error

Error using spinw/spinwave (line 831)
Hamiltonian matrix is not positive definite, probably the
magnetic structure is wrong! For approximate diagonalization try
the param.hermit=false option

Recall I saw this error previously for the supercell commensurate calculation and we thought maybe the supercell structure was subtly different - hence only doing hermit=False for the supercell calc in the script.

However if I enable mex then the incommensurate calc executes fine with hermit=True - hence the profile results saved here https://github.com/SpinW/performance/tree/main/PCWIN64_2022b_SpinW1634)

Is this expected? The non-mex chol function returns posDef but chol_omp doesn't return this - yet the doc string does say it will throw an error if the matrix is not positive definite. Maybe they use different tolerances?

UPDATE:
chol_omp does seem to throw an error for non positive definite matrices (at least in this case):

M = ones(2);
Romp = chol_omp(M)

UPDATE 2:
For the supercell calc. with mex enabled and hermit=True chol_omp does indeed throw an error
The input matrix is not positive definite

Originally posted by @RichardWaiteSTFC in #124 (comment)