CTSM runoff remapping issue again

We found incorrect runoff mapping from LND to ROF again. This powerpoint shows the problem
[cmep_history.pptx](https://github.com/ESCOMP/mizuRoute/files/12423309/cmep_history.pptx)

**What we did:**
- turn on cpl history file output and run CTSM-mizuroute with f09_f09_rHDMA (no lake) grid and f09_f09_rHDMAlk (including lake polygon) grid.
- run one year (1960)
- computed average runoff (`rofExp_Flrl_rofsub`+`rofExp_Flrl_rofsur`+`rofExp_Flrl_rofgwl`) in coupler history file to map them on mizuRoute grid and also map `Med_frac_rof_lfrac`

**What we found:**
- coupler runoff on rHDMAlk grid has severer problem than rHDMA which is better but not perfect.
- `Med_frac_rof_lfrac` spatial pattern does not look correct and there is also >1 fraction.
- It seems like regridding from LND to ROF itself (weighted average of CTSM runoff) is correctly done, but normalization by land fraction mess up the final runoff values.

**Additional information:**

mizuRoute uses pre-computed mapping file, not internally computed using ROF mesh and LND mesh.
```
LND2ROF_FMAPNAME: /glade/work/mizukami/py_mapping/mapping_conversion/data/map_fv0.9x1.25_TO_HDMA-lake_cdf5.nc
ROF2LND_FMAPNAME: /glade/work/mizukami/py_mapping/mapping_conversion/data/map_HDMA-lake_TO_fv0.9x1.25_cdf5.nc
```