Incremental Code Generation for Large Schemas

**Problem**

For large GraphQL schemas (1000+ types), gqlgen regenerates all code on every run, even when only a subset of schema files have changed. This leads to:

Unnecessary file rewrites - Files are rewritten even when content is identical, invalidating Go's build cache
Long rebuild times - Go recompiles the entire generated package (~1M+ lines) because file mtimes changed
Poor developer experience - Small schema changes trigger full multiple-minute rebuilds

**Proposed Solution**

Build a dependency graph tracking which types are defined in which schema files. When specific schemas change, compute the transitive closure of affected schemas and only regenerate those files.