Bump version.

Fix unexpected cpp output on specific machines.

Fix redundant loop when using restart inversion.

Automatically transfer data between host and device.

Use half as the sloppy precision in MultiCG when residues are close.

Remove odd-odd preconditioned situation in actions.

Fix a bug in multigrid solver.

Bug fix.

Speedup HISQ RHMC by using SINGLE in MultiCG.

Skip loading gauge field if it has been loaded.

Use the with syntax to manage gauge fields loaded by QUDA.