Add in-process no-fork runtime for Provider OCI

[kousalya-kethavath]

Description of your changes

This PR adds no-fork execution support for Crossplane Provider OCI using the public Terraform Provider OCI package.

Key changes include:

1. 4b7b5f0 : Add public-package OCI v9 no-fork runtime
   • Integrates the Terraform Provider OCI public Go package.
   • Adds in-process SDKv2 and Plugin Framework provider construction.
   • Adds no-fork configuration, routing, server-side apply support, and runtime resource registration.
   • Generates cluster-scoped and namespaced runtime schemes and service-provider entry points.
   • Adds tests for generation, provider configuration, routing, and server-side apply behavior.
2. 421b16f : Remove Terraform CLI from provider runtime image
   • Removes the Terraform CLI binary and Terraform CLI configuration from the provider runtime image.
   • Keeps the no-fork service runtime limited to the compiled Go provider process.
3. c86c2e7 : Optimize no-fork provider package publishing
   • Adds reusable Go build-cache warming for service packages.
   • Improves package batching and multi-architecture publishing.
   • Reduces repeated compilation and manages runner disk usage during package publication.
4. b1b823a : Use selective lazy OCI provider construction
   • Constructs the SDKv2 provider only when the reconciled resource requires it.
   • Uses service-scoped Terraform resource inventories instead of eagerly constructing the full provider.
   • Preserves the Plugin Framework execution path for applicable resources.
5. 23ea29a : Honor provider meta cache size in service runtimes
   • Propagates --provider-meta-cache-size to generated service-provider binaries.
   • Uses a bounded, provider-instance-owned cache rather than shared global state.
   • Supports concurrent initialization coalescing, configuration changes, and LRU eviction.
   • Adds regression and concurrency tests for cache behavior.
6. edf1fe9 : Add no-fork runtime metrics collector
   • Exposes Prometheus metrics for cache hits, misses, initialization results and duration, waiting callers, entries, capacity, inflight initialization, and evictions.
   • Avoids ProviderConfig identifiers or credential-derived metric labels.
   • Adds a reusable collector for pod resource usage, reconciliation state, restarts, process evidence, controller metrics, and no-fork cache metrics.
   • Adds focused metrics and cache tests.

The PR currently consumes Terraform Provider OCI’s feature/crossplane-support branch containing the required public-package and in-process changes.

I have:

• Read and followed Crossplane's contribution process.
• Run make generate and committed the results.
• Not made manual changes to generated files, and verified this with make check-diff.

How has this code been tested

The no-fork provider was validated through correctness, smoke, upgrade/rollback, authentication, provider-isolation, restart, and scale testing. Tests covered cluster-scoped and namespaced resources using API-key credentials, Instance Principal, and OKE Workload Identity.

Resources reconciled successfully without unexpected drift or recreation. In scale testing, no-fork reconciled all 100 resources with zero provider restarts and substantially lower peak memory than the Terraform CLI-based provider.