Use int64 for GitHub database IDs

Closes #9247

Problem

Several structs in the CLI use Go's int type for GitHub database ID fields. Go's int is platform-dependent and could be 32-bit on some architectures, while GitHub internally uses 64-bit integers for database IDs. The REST API OpenAPI spec explicitly declares many resource IDs as integer with format: int64 (e.g. repositories, GPG keys, SSH keys), and the GraphQL API has migrated some databaseId fields from Int to BigInt.

Changes

Updated all struct fields representing GitHub database IDs from int to int64, along with their downstream usage:

• Codespaces: Repository.ID, CreateCodespaceParams.RepositoryID, startCreateRequest.RepositoryID, plus interface/mock/test cascading
• Rulesets: RulesetGraphQL.DatabaseId, RulesetREST.Id, BypassActors.ActorId, RulesetRule.RulesetId, and the skills rulesetsResponse.ID
• Keys: deployKey.ID, gpgKey.ID, sshKey.ID
• Cache: Cache.Id
• Autolinks: Autolink.ID
• Agent tasks: JobActor.ID, JobPullRequest.ID

All strconv.Itoa() calls on these fields were updated to strconv.FormatInt(x, 10).

GraphQL response structs were also audited - only RulesetGraphQL.DatabaseId needed updating. Others were already correct (GitHubUser.DatabaseID is int64, FullDatabaseID is string for GraphQL BigInt).

Lint check

Added a custom go/analysis linter (cmd/idtype-checker) that flags struct fields with ID-like names or JSON tags that use int instead of int64. Integrated into make lint to prevent regressions.

Validation

• go build ./... passes
• All affected package tests pass
• make lint reports 0 issues
• Linter correctly catches all 27 pre-fix violations with zero false positives