Summary

Fixes the supra-adjacency docs mismatch reported in #1262 and addresses the underlying multiplex generator behavior that made the documented expectation unreliable.

This PR does two things:

1. Fixes the documentation to use the correct generator for a classic multiplex supra-adjacency example.
2. Fixes random_multiplex_ER() so it always materializes all node replicas across all layers, including isolated ones.

Problem

The docs example in Section 3.8 expected:

• 100 nodes
• 3 layers
• supra-adjacency shape (300, 300)

but used random_multilayer_ER(...), which generates a general multilayer network where each base node is assigned to a single layer. In that case, the supra-adjacency size reflects
the number of node-layer pairs that actually exist, which is typically n, not n * l.

Separately, random_multiplex_ER(...) did not explicitly add isolated replicas, so even in multiplex mode the supra-adjacency shape was not guaranteed to be exactly n * l when some
nodes had no incident edges in a layer.

Root cause

• Docs issue: the example used random_multilayer_ER(...) where random_multiplex_ER(...) was intended.
• Implementation issue: random_multiplex_ER(...) only added nodes implicitly through edges, so isolated node-layer replicas could be missing.

Changes

Code

• Update random_multiplex_ER() to add all (node, layer) replicas before adding edges.
• Make multiplex supra-matrix encoding robust for empty graphs.
• Use stable base-node ordering during multiplex encoding so layer blocks align consistently.

Docs

• Update the supra-adjacency example in the book to use random_multiplex_ER(...).
• Clarify the semantic difference between:
  • random_multilayer_ER(...) → general multilayer, present replicas only
  • random_multiplex_ER(...) → strict multiplex, all n * l replicas

Tests

• Add regression coverage for:
  • random_multilayer_ER(...) producing a supra shape based on present replicas
  • random_multiplex_ER(...) preserving all replicas even when p=0

Expected behavior after this PR

• random_multilayer_ER(100, 3, p) produces a supra-adjacency matrix sized by the actual node-layer pairs present.
• random_multiplex_ER(100, 3, p) produces a supra-adjacency matrix of shape (300, 300).

Validation

Direct sanity checks confirm:

• random_multilayer_ER(12, 3, 0.4) → (12, 12)
• random_multiplex_ER(8, 4, 0.0) → (32, 32)
• random_multiplex_ER(100, 3, 0.05) → (300, 300)

Closes

Closes #1262