Flow .NET Servicing versions----#### AI description  (iteration 1)#### PR ClassificationThis PR performs a servicing update by bumping .NET dependency versions and refining CI pipeline configurations for stable releases.#### PR SummaryThe changes update dependency versions from 9.0.6 to 9.0.7 (and corresponding LTS versions) and adjust build settings to support release stability while streamlining internal feed configurations.- `eng/Version.Details.xml`: Updated multiple dependency version numbers and SHA values from 9.0.6 to 9.0.7.- `eng/Versions.props`: Bumped version properties (including LTS versions from 8.0.17 to 8.0.18) and enabled stable release settings by setting package stabilization to true and DotNetFinalVersionKind to release.- `NuGet.config`: Modified package source settings by adding new internal feed mappings and removing preexisting package source mapping blocks.- `azure-pipelines.yml` & `eng/pipelines/templates/BuildAndTest.yml`: Removed the CodeCoverage stage and added tasks for setting up private feed credentials, with integration tests commented out due to authentication requirements.- `Directory.Build.props`: Suppressed NU1507 warnings to accommodate internal feeds without package source mapping.<!-- GitOpsUserAgent=GitOps.Apps.Server.pullrequestcopilot -->

* Add DelegatingAIFunctionTo simplify scenarios where someone wants to augment an existing AIFunction's behavior, tweak what one of its properties returns, etc.* Address PR feedback

The conventions we're pushing for on NuGet are:1. `PackAsTool`2. `McpServer` package type (in addition to the default `DotnetTool`)3. Embed server.jsonThis provides an `mcpserver` template as part of `Microsoft.Extensions.AI.Templates`. This currently only covers a local MCP server, with stdio transport.

We've had a bunch of requests to be able to customize how function invocation is handled, and while it's already possible today by deriving from FunctionInvokingChatClient and overriding its InvokeFunctionAsync, there's a lot of ceremony involved in that. By having a property on the client instance, that behavior can instead be configured as part of a UseFunctionInvocation call.

* Add reporting tests that show NLP results.* Cleanup analyzer errors.* Add global tags for NLP* Add more precision to the evaluator timing* More tags* Add another partial match test

* Enable specifying "strict" for OpenAI clients via ChatOptions* Address PR feedback

* AIFunctionFactory: tolerate JSON string function parameters.* Add debug assertion.* Update src/Libraries/Microsoft.Extensions.AI.Abstractions/Functions/AIFunctionFactory.csCo-authored-by: Stephen Toub <stoub@microsoft.com>* Add regex-based JSON string recognition and add more tests.---------Co-authored-by: Stephen Toub <stoub@microsoft.com>

…ns for 9.7Pure cherry-picks

Clean cherry-picks