fix: update oasdiff to v1.14.0 and migrate from oasdiff/kin-openapi to getkin/kin-openapi

[Copilot]

Fixes the build failures caused by upgrading github.com/oasdiff/oasdiff from v1.13.1 to v1.14.0 (see dependabot PR #1208).

oasdiff v1.14.0 introduced two breaking changes:

1. Dependency migration: Migrated from its own github.com/oasdiff/kin-openapi fork to the upstream github.com/getkin/kin-openapi, causing type incompatibilities across ~55 Go files in tools/cli and 6 files in tools/mcp-server.
2. Message format change: Changed GetUncolorizedText() output from single-quoted values ('value') to backtick-quoted values (`value`), and added the HTTP status code as an extra argument to several response-related message templates.

Changes:

• tools/cli/go.mod: Replace github.com/oasdiff/kin-openapi v0.136.10 with github.com/getkin/kin-openapi v0.135.0 (the version used by oasdiff v1.14.0) and bump oasdiff to v1.14.0
• tools/mcp-server/go.mod: Replace github.com/oasdiff/kin-openapi v0.136.10 with github.com/getkin/kin-openapi v0.135.0 as a direct dependency
• ~61 Go source/test files: Replace all imports of github.com/oasdiff/kin-openapi/openapi3 with github.com/getkin/kin-openapi/openapi3 and apply gci import formatting across both tools/cli and tools/mcp-server
• tools/cli/go.sum and tools/mcp-server/go.sum: Updated via go mod tidy
• .gitignore: Added tools/go.work.sum to prevent the auto-generated workspace checksum file from being accidentally committed
• squash.go: Updated identifierRegex from '([^']*)' to `([^`]*)` to match backtick-quoted values; replaced the "" placeholder with \x01 (ASCII SOH) to avoid false regex matches during the two-phase text construction (phase 1 marks substitutions with \x01<value>\x01; phase 2 converts \x01 → ' for the final output format)
• enum.go: Updated expectedNumberOfValues for response-property-enum-value-added/removed and response-write-only-property-enum-value-added (2→3), and response-mediatype-enum-value-removed (1→2, squashIdx 0→1) to handle the new status code and newly-quoted media type arguments; wrapped long comment examples across two lines to satisfy the lll linter (150-character limit)
• field.go: Added the new expectedNumberOfValues parameter to squashFieldChanged — set to 1 for request handlers (which have a single backtick-quoted value: the property name) and 2 for response handlers (which now include the HTTP status code as a second backtick-quoted value in oasdiff v1.14.0)
• Unit tests (squash_test.go, enum_test.go, field_test.go): Updated input texts to use backtick format; expected outputs continue to use single-quote format (squash output) with status codes now included; added //nolint:lll //Test string annotations to test string literals that inherently exceed the 150-character line limit, matching the existing pattern used elsewhere in those files
• E2e test data: Regenerated all 5 test data directories (new-api-version, same-api-version, new-api-preview-version, new-upcoming-version, rename-api-version) using the updated CLI binary with --run-date 2025-06-15, matching the fixed date used by the e2e test

Jira ticket: CLOUDP-#

Checklist

• I have signed the MongoDB CLA
• I have added tests that prove my fix is effective or that my feature works

Changes to Spectral

• I have read the README file for Spectral Updates

Further comments

Validation:

• go build ./... ✅ (both tools/cli and tools/mcp-server)
• go test ./internal/... ✅ (all 25 test packages pass)
• gci import formatting applied to all changed Go files ✅
• lll lint violations resolved in enum.go, enum_test.go, and field_test.go ✅
• E2e test data regenerated with updated CLI binary and correct --run-date 2025-06-15 flag ✅
• TestChangelog/Generate_Changelog_with_Upcoming_API_Version e2e test passes ✅

[andreaangiolillo]

@copilot Why do we need this file?

[Copilot]

This file was accidentally generated when running go mod tidy in a workspace context. It's not needed — removed in 825e855.

[andreaangiolillo]

@copilot the lint test if failing. Could you fix it? logs: https://github.com/mongodb/openapi/actions/runs/24454623724/job/71452311418?pr=1228

[Copilot]

Fixed in cd3d20f. The lll linter was flagging 8 lines exceeding 150 characters:

• enum.go lines 9, 22, 48: wrapped long comment examples across two lines
• enum_test.go lines 22, 29, 185, 241 and field_test.go line 192: added //nolint:lll //Test string annotations, matching the existing pattern already used in those files

[andreaangiolillo]

@copilot Why did you introduced this placeholder? what are the advantages and disadvantages of using it? why this was not defined before?

[Copilot]

The placeholder (\x01, ASCII SOH) was introduced because replaceOnlyFirstOccurrence needs two-phase text construction:

1. Phase 1: It replaces each backtick-quoted value in the template with \x01<value>\x01 (e.g. `DUBLIN_IRL` → \x01DUBLIN_IRL\x01)
2. Phase 2: After all values are substituted, a final strings.ReplaceAll converts every \x01 → ' to produce the output format (e.g. 'DUBLIN_IRL')

Why not use "" (empty string)?
Before oasdiff v1.14.0, the input messages used single-quoted values (e.g. 'DUBLIN_IRL'). The old placeholder was "", which simply wrapped the replacement value in empty strings — harmless because single quotes couldn't appear inside another single-quoted value.

With v1.14.0, the input now uses backticks (e.g. `DUBLIN_IRL`). If we still used "" as the placeholder during Phase 1, any replacement value that itself contains a backtick (e.g. a path like `/items/region`) would accidentally match the identifierRegex again in subsequent iterations of ReplaceAllStringFunc, causing double-substitution.

\x01 (SOH) is safe because it's a control character that oasdiff never includes in its message text, so it can't create a false regex match during Phase 1.

[andreaangiolillo]

LGTM