feat: add --ignore-link flag to skip URLs in link validation

[blva]

Closes #68

Changes

Adds an --ignore-link flag to check and validate links that skips any URL containing a matching substring (case-insensitive). The flag is repeatable and also accepts comma-separated patterns, consistent with --only, --skip, and --allow-dirs.

# Skip a private org's repos
skill-validator validate links --ignore-link=github.com/myorg .

# Multiple patterns: repeated or comma-separated (equivalent)
skill-validator check --ignore-link=github.com/myorg --ignore-link=localhost .
skill-validator check --ignore-link=github.com/myorg,localhost .

Implementation

• links.Options — new struct with IgnorePatterns []string
• links.CheckLinks — updated signature to accept Options; filters ignored URLs before the concurrent check loop
• links.isIgnored — case-insensitive substring helper
• orchestrate.Options — adds LinksOpts links.Options, threaded through RunAllChecks and RunLinkChecks
• cmd/check.go and cmd/validate_links.go — expose --ignore-link flag, wire into options
• README.md — documents the flag in both the validate links and check command sections, and in the "Link validation" what-it-checks section

Tests

• links.TestIsIgnored — 8 table-driven cases: empty patterns, exact match, case-insensitive, partial substring, no-match, first/second of multiple patterns, none matching
• links.TestCheckLinks_IgnorePatterns — 4 cases using httptest.NewServer: ignored URL skipped, non-ignored URL still checked, mixed body, case-insensitive matching
• orchestrate.TestRunLinkChecks_WithIgnorePatterns — uses the invalid-skill fixture (has https://httpstat.us/404); asserts 0 errors when ignored, errors when not
• orchestrate.TestRunAllChecks_WithIgnoreLinks — same fixture; asserts LinksOpts propagates through RunAllChecks
• cmd.TestSliceFlags — 3 integration tests exercising the repeated flag form, comma-separated form, and the flag on check --only=links; all run against the compiled binary

[dacharyc]

Thanks for this, the implementation is clean and the test coverage is solid! A few things I'd like to address before merge:

Bug: empty pattern matches everything

In Go, strings.Contains(s, "") returns true for any s. This means an accidental empty pattern silently disables all link validation:

# All of these skip every link with no warning:
skill-validator validate links --ignore-link= .
skill-validator validate links --ignore-link=,github.com/myorg .
skill-validator validate links --ignore-link="github.com/myorg," .

isIgnored should skip empty strings in the patterns slice. Would also be good to add a test case for this.

Visibility into skipped links

Right now ignored links are silently dropped. If someone misconfigures a pattern that's too broad (e.g. --ignore-link=github.com instead of --ignore-link=github.com/myorg), they'd have no way to know they're accidentally skipping legitimate links.

We do something similar for 403 responses at links/check.go:122, where we emit an info-level result rather than silently swallowing the ambiguity. Could we do the same here? Something like an rctx.Infof("%s (skipped — matches ignore pattern)", url) per link, or a summary count. Either way, the user should see what was skipped.

Substring matching breadth

Minor, but worth noting: --ignore-link=example.com would also match notexample.com or example.com.evil.org. For the motivating use cases (org paths, localhost) this is unlikely to matter, but could we note this in the flag help text or README? Something like "the pattern is matched as a substring, not a domain boundary" so users aren't surprised.