Subagent-scoped custom tools, context compaction lifecycle hooks, and universal JSON schema normalization

The 0.1.15 release introduces subagent-exclusive tool scoping to reduce context overhead, adds an on_compaction lifecycle hook for observing context checkpointing, expands platform compatibility with Alpine Linux musl wheels, resolves workspace path normalization and custom Vertex endpoint routing, and adds universal JSON Schema normalization for custom Python tools when targeting local OpenAI-compatible LLM endpoints.

• Subagent-Scoped Custom Tools: Custom tools can now be registered directly on subagents without requiring registration on the root agent, strictly isolating them to the subagent's context and reducing root context tokens.
• Context Compaction Lifecycle Hook: Improved support for the on_compaction lifecycle hook to accurately capture compaction events and summaries when long-running conversations trigger checkpointing.
• Custom Base URL & Secure Vertex Proxy Support: VertexEndpoint now supports routing requests to custom base_url reverse proxies or enterprise gateways without leaking ambient Google Cloud Application Default Credentials (ADC) OAuth tokens or conflicting with project/location configurations.
• Universal JSON Schema Normalization for Custom Tools: Custom Python tools now produce standard OpenAPI / JSON Schema compliant parameter definitions with lowercase types and camelCase combiners, eliminating schema errors when connecting to local OpenAI-compatible engines such as Ollama, LM Studio, or vLLM.
• from_bytes Helper Export: Exported the from_bytes helper in top-level google.antigravity alongside from_file for creating binary and multimodal content payloads.
• PEP 656 musllinux Wheel Support: Added musllinux_1_1 wheel platform tags for x86_64 and aarch64 architectures, enabling direct installation via pip on Alpine Linux containers.
• Isolated Harness Environment Configuration: Added per-connection environment dictionary resolution for ANTIGRAVITY_HARNESS_PATH, avoiding mutation of global os.environ.
• Tool Call Metadata Preservation: Preserved id, step_id, and server_name metadata across all ToolResult executions, including batch calls, errors, and unknown tools.
• Relative Workspace Path Normalization: LocalAgentConfig(workspaces=...) now resolves relative directory paths and tilde (~) expansions against the current working directory (os.getcwd()).
• Local OpenAI Tool Schema Validation: Fixed HTTP 400 "Invalid discriminator value" errors on local OpenAI endpoints by canonicalizing tool parameter schemas to standard JSON Schema.
• Vertex Custom Gateway Token Leaks: Fixed host GCP OAuth bearer token leakage and environment variable collisions when using custom base_url endpoints with VertexEndpoint.
• Subagent Custom Tool Routing: Fixed tool dispatch failures when subagents defined tools not registered on the root agent.
• UsageMetadata Service Tier Loss: Fixed UsageMetadata.sub dropping the service_tier field when calculating per-turn usage deltas.
• WebSocket Deprecation Warnings: Resolved DeprecationWarning exceptions when reading WebSocket close codes across varying websockets library versions.
• OpenTelemetry Optional Dependency Guard: Fixed test collection failures on minimal environments lacking opentelemetry.sdk.
• Deprecated TriggerDelivery Removal: Removed the unused TriggerDelivery enum from types.py.