add small fix to HLLD solver from Athena++

[BenWibking]

Describe the bug
One of the degeneracy checks in the HLLD solver should not be used. See: PrincetonUniversity/athena#620

Additional context
From PrincetonUniversity/athena#620:

It turned out that we should not remove the double-star states even when the normal magnetic field bx is very weak. If this branch is selected when bx is actually non-zero, it may return unphysical flux. In fact, this check is not mentioned in Miyoshi & Kusano 2005 but already existed in the old Athena,

I do not think this has affected many people, as this is problematic only when the magnetic fields are very weak and the flow is not smooth. The probability that this branch is selected is very low in a realistic flow.