Implement candidate selection probability style inv_pop_f

[wanghan-iapcm]

This style

1. calculate the population of frames as a function of force model deviation, i.e. P(md_f), in range [ level_f_lo, level_f_hi ]
2. select the candidate with probability 1./P(md_f).

[codecov[bot]]

Codecov Report

Patch coverage: 90.19% and project coverage change: +0.03% :tada:

Comparison is base (a6f0fec) 88.05% compared to head (b3b7746) 88.09%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #157 +/- ## ========================================== + Coverage 88.05% 88.09% +0.03% ========================================== Files 80 80 Lines 3961 4007 +46 ========================================== + Hits 3488 3530 +42 - Misses 473 477 +4 ``` | [Files Changed](https://app.codecov.io/gh/deepmodeling/dpgen2/pull/157?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=deepmodeling) | Coverage Δ | | |---|---|---| | [dpgen2/exploration/report/report\_adaptive\_lower.py](https://app.codecov.io/gh/deepmodeling/dpgen2/pull/157?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=deepmodeling#diff-ZHBnZW4yL2V4cGxvcmF0aW9uL3JlcG9ydC9yZXBvcnRfYWRhcHRpdmVfbG93ZXIucHk=) | `95.97% <88.88%> (-1.43%)` | :arrow_down: | | [...en2/exploration/report/report\_trust\_levels\_base.py](https://app.codecov.io/gh/deepmodeling/dpgen2/pull/157?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=deepmodeling#diff-ZHBnZW4yL2V4cGxvcmF0aW9uL3JlcG9ydC9yZXBvcnRfdHJ1c3RfbGV2ZWxzX2Jhc2UucHk=) | `96.52% <100.00%> (+0.03%)` | :arrow_up: |

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.