Report per-file license text coverage

[abhinavgautam01]

Closes #8

Summary

Add per-file license-text coverage to the scanner's JSON output so consumers can determine whether a file should be treated as complete attribution text.

Each file record now includes:

"license_text_coverage": 95.4

The value is the percentage of decoded file bytes covered by the union of text- and notice-kind match spans.

Implementation

• Add LicenseTextCoverage to fileRecord
• Combine overlapping and adjacent match spans without double-counting
• Include only text and notice matches
• Ignore reference, tag, introduction, and clue matches
• Clamp spans to the decoded input bounds
• Calculate coverage before offsets are remapped to the original encoding
• Return 0 when no qualifying matches are present

Calculating coverage against decoded bytes keeps the result consistent across UTF-8, UTF-8 BOM, UTF-16LE, UTF-16BE and Latin-1 files.

The JSON schema remains version 1 because this is an additive field and schema 1 already requires consumers to accept unknown fields.

Testing

Added coverage for:

• Overlapping text and notice spans
• Exclusion of unrelated match kinds
• Invalid and out-of-bounds spans
• Empty and unmatched files
• UTF-8, UTF-8 BOM, UTF-16LE, UTF-16BE and Latin-1 inputs
• Presence of license_text_coverage in JSON output

Validation

• gofmt
• git diff --check
• go build ./...
• go vet ./...
• go test -count=1 ./...
• go test -race -count=1 ./...
• golangci-lint v2.12.2 — 0 issues