[smart-log] perf: optimize stack discovery with batch fetch

[luqven]

Summary

Optimize stack discovery by replacing sequential API calls with a batch-fetch strategy. This dramatically improves performance for the gh-stack log and gh-stack status commands.

Problem

The previous implementation made O(N+M) sequential API calls to walk the PR chain:

• Walk UP: fetch_pr_by_head() for each ancestor
• Walk DOWN: fetch_prs_by_base() for each descendant

For a 6-PR stack, this meant ~12 HTTP requests with network latency (~100-500ms each), resulting in 6+ seconds of wait time.

Solution

Batch fetch all open PRs first, then walk the chain in-memory:

1. Single paginated API call to fetch all open PRs
2. Build in-memory PrIndex for fast head/base lookups
3. Walk the chain without any additional API calls

Changes

Modified: src/api/stack.rs

• Add pagination to fetch_all_open_prs() - supports up to 1000 PRs (10 pages)
• Add PrIndex struct for O(1) lookup by head or base branch
• Refactor discover_stack() to use batch fetch + in-memory walk
• Extract discover_stack_from_index() for pure in-memory operation
• Extract group_into_stacks() for code reuse

Performance

Scenario	Before	After
6-PR stack	~12 API calls (~6s)	1-2 API calls (~2s)
Any stack size	O(N) calls	O(1) calls

Enterprise Support

• Pagination handles repositories with 100+ open PRs
• Capped at 1000 PRs (10 pages) for safety and to avoid runaway requests

Tests

11 new tests added:

• test_pr_index_get_by_head / get_by_base
• test_discover_stack_from_index_* (4 scenarios)
• test_group_into_stacks_* (3 scenarios)
• test_fetch_all_open_prs_pagination
• test_discover_stack_batch_fetch

Stacked PR Chain: smart-log

PR	Title	Merges Into
#35	feat: add stack discovery API	-
#36	feat: add rate limit retry logic	#35
#37	feat: add identifier detection and prompts	#36
#38	feat: add gh CLI helpers	#37
#39	feat: make log and status identifier optional with branch inference	#38
#40	docs: update log and status documentation for smart defaults	#39
#41	👉perf: optimize stack discovery with batch fetch	#40
#42	perf: parallelize status check fetches	#41

[luqven]

Landed via #42