OEI performance issue in QM/MM simulations

[Madu86]

When the latest QUICK master branch is coupled with Sander for QM/MM simulations, the performance of OEI gradient calculation is surprisingly low. This has to be investigated and fixed before the Amber release.

QM/MM tests (QM2 B3LYP/def2-svp): AmberTools23

| INITIAL GUESS TIME = 2.826331000( 3.71%)
| DFT GRID OPERATIONS = 1.321008000( 1.73%)
| TOTAL SCF TIME = 30.983988000( 40.62%)
| TOTAL OP TIME = 29.002852000( 38.02%)
| TOTAL 1e TIME = 0.972896000( 1.28%)
| TOTAL 2e TIME = 18.587176000( 24.37%)
| TOTAL EXC TIME = 8.166470000( 10.71%)
| TOTAL DII TIME = 1.933969000( 2.54%)
| TOTAL DIAG TIME = 0.481339000( 0.63%)
| TOTAL GRADIENT TIME = 36.970580000( 48.47%)
| TOTAL 1e GRADIENT TIME = 4.790592000( 8.47%)
| TOTAL 2e GRADIENT TIME = 26.900787000(35.27%)
| TOTAL EXC GRADIENT TIME = 3.545164000( 4.65%)
| TOTAL TIME = 76.273830000

| Job cpu time: 0 days 0 hours 1 minutes 16.3 seconds.

Development branch:

| INITIAL GUESS TIME = 2.871584000( 1.79%)
| DFT GRID OPERATIONS = 1.353214000( 0.84%)
| TOTAL SCF TIME = 52.377066000( 32.64%)
| TOTAL OP TIME = 50.557895000( 31.50%)
| TOTAL 1e TIME = 11.687946000( 7.28%)
| TOTAL 2e TIME = 18.766085000( 11.69%)
| TOTAL EXC TIME = 8.162163000( 5.09%)
| TOTAL DII TIME = 1.778441000( 1.11%)
| TOTAL DIAG TIME = 0.476679000( 0.30%)
| TOTAL GRADIENT TIME = 110.775731000( 69.03%)
| TOTAL 1e GRADIENT TIME = 4.678534000(46.59%)
| TOTAL 2e GRADIENT TIME = 32.427885000(20.21%)
| TOTAL EXC GRADIENT TIME = 3.548908000( 2.21%)
| TOTAL TIME = 160.477502000

| Job cpu time: 0 days 0 hours 2 minutes 40.5 seconds.