Rename d_mesh_static dimension to follow canonical naming convention

[joeloskarsson]

Background

PR #635 added a "Canonical dimension names" list to the README, including:

• d_mesh_static - number of static features per mesh node

In review of PR #635 it was noted that d_mesh_static doesn't
match the naming pattern of the other canonical names: num_* for counts (num_state_vars, num_forcing_vars, num_grid_nodes,
num_mesh_nodes) and *_dim for feature sizes (hidden_dim, input_dim). The d_-prefix follows neither.

Proposal

Rename d_mesh_static to num_mesh_static_vars, matching the num_* family.

Scope

The name predates #635 and also appears in code comments/docstrings, which should be updated together:

• README.md — canonical dimension names list (added in feat: add optional boundary datastore support #635)

• neural-lam/neural_lam/utils.py

  Line 328 in 10e93a6

  ) # List of (N_mesh[l], d_mesh_static)

• neural-lam/neural_lam/create_graph.py

  Line 285 in 10e93a6

  - mesh_features.pt list of [N_mesh_nodes_level, d_mesh_static],

• neural-lam/neural_lam/create_graph.py

  Lines 299 to 301 in 10e93a6

  	d_mesh_static:
  	number of static features per mesh node (currently
  	d_mesh_static==2, for x and y)

Separately, surrounding comments use ad-hoc node-count names (N_mesh, N_mesh_nodes_level) instead of the canonical
num_mesh_nodes. Would be good to also fix

[AkankshaRaj07]

@joeloskarsson Hi! I'm interested in working on this issue. I'll update the canonical naming in the README and comments/docstrings as described. Could you please assign it to me?