Refactor: Replace nlm_save in DeePKS by HContainer object phialpha.

[ErjieWu]

Linked Issue

Fix #5744 , set_size() is defined in BaseMatrix now and will be called when set_size() is called in AtomPair.

Unit Tests and/or Case Tests for my changes

• Change two reference results of integrate test in DeePKS for stress alignment. Before this update, the stress calculation in the DeePKS part was processed using a lower triangular matrix for gamma-only condition. In the new code, it has been changed to an upper triangular matrix to be consistent with other stress calculation codes within ABACUS. There are certain numerical differences in the results after the change, so the result.ref file has been updated here.
• Change the name of psialpha into phialpha in integrate test.

What's changed?

• Use Hcontainer to build a new variable phialpha in LCAO_Deepks to replace nlm_save and nlm_save_k, overlap between projected orbitals and basis are now saved in same form for both gamma-only case and multi-k case.
• Add set_size() function in hamilt::BaseMatrix and call it in AtomPair to avoid wrong results.
• Add cal_stress value check for deepks_out_unittest=1.
• Change contributeHR() in deepks_lcao into template.
• Change all the name psi in DeePKS into phi since they are actually NAO basis.

Notice

• The output file deepks_psialpha.npy is also changed into deepks_phialpha.npy, this may cause error for reading files when using deepks-kit (deepks-kit should be changed later as well, or users may need to change the file name into old version by hand).