Update computational homogenization tutorial to not use legacy PeriodicDirichlet?

[KnutAM]

Noticed when reviewing #1224 that the usage in the tutorial isn't documented in the API, and this also gives a warning during docs-build...

[fredrikekre]

We need a way to query for a corner dof to lock to 0 I think.

[lijas]

I think if we let the user handle the locking of the corner nodes with Dirichlet, then we might get incompatible constraints. I think the legacy code is needed to handle this correctly