doc: harmonic calculations from a phono3py-yaml file

[VariableYT]

Addresses #592.

Documents that phonopy.load() on a phono3py-yaml file returns a Phonopy
instance built on supercell_matrix rather than phonon_supercell_matrix,
and does not carry the phonon dataset, as intended behaviour rather than a
bug.

Includes the errors it produces (Displacement dataset is not set,
Dynamical matrix has not yet built) so the section is findable by search,
and the recommended route using phono3py.load and phonon_dataset, which
is the code you gave in #592.

Added as the last subsection of "Use of different supercell dimensions for
2nd and 3rd order FCs", since that section already covers
phonon_supercell_matrix. Uses the AlN-LDA example file already used
elsewhere in the document.

Happy for this to be rewritten or closed if you would rather write it
yourself or put it somewhere else.

[atztogo]

Thank you for the PR updating the docs! I'd like to write this documentation myself, so I'll close this one, but I really appreciate you taking the time.

[VariableYT]

No problem at all, and thanks for the detailed answers throughout! Glad the
behavior will be documented.

[atztogo]

@VariableYT, sorry being very late to work on this. I have updated the document. I revised some old information and added a section for displacement dataset, too.

[VariableYT]

@atztogo Thanks for coming back to this. The silent-truncation note for type-I datasets is exactly what would have saved me the diagnosis time in July. Appreciate it.