Pre-compute spatial graph statistics during Data initialization

[orien]

Context

The spatial entropy calculation uses graph statistics (average degree and starting positions) to compute password strength for keyboard patterns. Previously, these statistics were calculated on-demand for every spatial match by iterating through the adjacency graph data, performing map/compact/inject operations each time.

Since adjacency graphs are immutable after loading, these statistics can be computed once during initialisation and reused, eliminating redundant calculations.

Changes

Data class (lib/zxcvbn/data.rb):

• Added compute_graph_stats private method that pre-computes average degree and starting positions for all adjacency graphs during initialisation
• Stores results in @graph_stats hash with structure: { graph_name => { average_degree:, starting_positions: } }
• Exposes statistics via graph_stats reader

Math module (lib/zxcvbn/math.rb):

• Updated average_degree_for_graph to retrieve pre-computed value from data.graph_stats
• Updated starting_positions_for_graph to retrieve pre-computed value from data.graph_stats
• Eliminated repeated map/compact/inject operations on graph data

Performance

Benchmark results (1000 iterations across 10 passwords with spatial patterns):

• Before: 0.097ms per password
• After: 0.088ms per password
• Improvement: 9.3% faster

The improvement comes from eliminating repeated iterations over adjacency graph data during entropy calculations. All 262 tests pass, confirming correctness is maintained.