perf(#34): hoist find_next_pending_index to eliminate O(N²) gas complexity in get_all_milestones_view

[d3vobed]

Summary

get_all_milestones_view called views::get_milestone_by_index for each milestone, and get_milestone_by_index invoked find_next_pending_index which linearly scanned every milestone to find the first non-released one. With N milestones this produced N + (N × N) ≈ N² storage reads.

Change

Hoist find_next_pending_index — compute it once at the top of get_all_milestones_view and inline the MilestoneView construction, reusing the precomputed value for is_next_pending:

Before	After
for i in 0..N → get_milestone_by_index → find_next_pending_index (full scan)	next_pending = find_next_pending_index(env) (1 scan)
	for i in 0..N → next_pending == i (O(1))
O(N²) storage reads	O(N) storage reads

The same fix can be applied to any other view function that iterates milestones and per-row calls find_next_pending_index.

Verification

• Behaviour unchanged — is_next_pending semantics are identical (same comparison, same value)
• All 142 tests pass (including existing snapshot fixtures for get_all_milestones)
• No public API changes — function signature and MilestoneView shape are untouched

Closes #34