[MCP Tool] Implement ability to specify outputSchema on tools

Implement support for specifying an outputSchema on MCP tool definitions. This allows tools to declare the JSON schema of their structured output, enabling clients to validate and better understand tool responses.

Background

The MCP specification allows tools to declare an output schema that describes the structure of the tool's return value. This is separate from the input schema (which describes the tool's parameters).

MCP Specification: https://modelcontextprotocol.io/specification/2025-06-18/server/tools#output-schema

Output Schema: Worker ↔ Host Contract (based off of this PR)

1. Worker side — configuration

The user calls the fluent builder:

builder.ConfigureMcpTool("MyTool")
    .WithOutputSchema("""{ "type": "object", "properties": { ... } }""");

WithOutputSchema validates the JSON string and stores it as McpOutputSchema on ToolOptions.

2. Worker side — metadata emission

During function metadata transformation, ResolveToolOutputSchemaExtension reads ToolOptions.OutputSchema and writes its validated JSON string onto McpParsedBinding.OutputSchema. McpBindingBuilder.Build() then injects it into the raw binding JSON under the key:

"outputSchema": "<validated json string>"

No flag (analogous to useWorkerInputSchema) is needed — if the key is present there is a schema; if absent, there isn't.

3. Host side — reading the contract

McpToolTriggerAttribute exposes OutputSchema (string?) populated from the raw binding JSON. McpToolTriggerBinding.GetOutputSchema() parses and validates it, requiring:

• Valid JSON
• Root must be type: "object" (validated via McpInputSchemaJsonUtilities.IsValidMcpToolJsonSchema)

The resulting JsonElement? is passed to McpToolListener, which advertises it as the tool's outputSchema in the MCP tools/list response.

Summary