Cherry-pick commits from `master`

[njzjz]

It's hard to merge devel with master since the difference is too large. So I suggest cherry-pick and force-push instead.

PRs since v1.3.0 are as follows. We needs to check if a PR has been patched to devel or is out-dated.

• bug fixing: not displaying input doc on github #312 (out-dated)
• Bug fixings and add notice for the consistency of TF versions.  #327 (out-dated)
• fix bug of nbor sorting #334
• fix an error in stress by ase interface #338 (cherry pick #338: fix an error in stress by ase interface  #964)
• update compiler support for Ampere architecture devices #354 (out-dated)
• fix bug of cuda-11 compilation #390 (out-dated)
• Fix lammps related bugs #392 (Refactorize the implementation of the neighbor list building. #391)
• fix bug of illegal device memory access #407
• fix bug of max_nbor_size error #424
• give the correct tensorflow version in the document #446 (out-dated, but needs to pay attention)
• Change bug issue template according to what Yuzhi proposed. #492 (Cherry pick #492 and #664: create issue templates #961)
• Do not find protobuf for python #535 (cherry pick #535: Do not find protobuf for python #963)
• only enable link what you use on GNU compilers #536 (cherry pick #536: only enable link what you use on GNU compilers #962)
• add more badges #564 (cherry pick #564: add more badges #965)
• add GPU UUID support for DP #620 (add UUID support for DP #621)
• append to out_file when LAMMPS restarts #641 (append to out_file when LAMMPS restarts #640)
• Update bug-report.md #664 (Change bug issue template according to what Yuzhi proposed. #492)
• fix class name in #641 #676 (out-dated)
• add aliases to Arguments #932 (add aliases to Arguments #933)
• fix a typo in #932 #934 (add aliases to Arguments #933)
• rm load_ckpt #936 (outdated)

[njzjz]

@denghuilu Can you make sure we can ignore #334, #407, and #424?

[denghuilu]

These three PRs can be safely ignored. #334 was out-dated, and #407,#427 have been patched into devel branch.