Metrics tool supports bandwidth usage (#34)

Improve get_metrics tools (#29)

- Gracefully handle a 400 when trying to request HTTP latency on a Hobby tier plan
- Support aggregating HTTP response counts
- Support filtering on path and host for HTTP counts and latencies
- Support HTTP latency quantile
- Support more metrics endpoints: CPU/mem limit and targets, which helps LLMs tune autoscaling
- A bit of code clean up (renaming variables, getting rid of some unneeded params and return values)


🤖 Generated with [Claude Code](https://claude.ai/code)


---------

Co-authored-by: Claude <[email protected]>

Adding ability to mcp server to pull service metrics (#28)

* Adding ability to mcp server to pull service metrics

* Addressing Previsou PR feedback from Ammar

* Fixing issues and adding ability for the MCP plugin to get instance count

* Add HTTP error and response time metrics support
- Add httperrors and responsetime metric types to MetricType constants
- Implement getHTTPErrorMetrics() for error breakdown by status code
- Implement getResponseTimeMetrics() for P95 latency percentiles
- Update tool description and validation to include new metric types
- Add comprehensive test coverage for new metric types
- Update README documentation with usage examples

Fix release version not being set (#14)

Remove --sensitive-info flag (#13)

Add missing .goreleaser.yml (#7)

Don't include sensitive information in MCP server response (#5)

Try to avoid pulling sensitive information into MCP host context.

Currently, the only sensitive info I can think of are environment variables and postgres/keyvalue connection strings.

Users can opt out of this behavior using the --sensitive-info flag if they're confident that their LLM will not train on this type of data.