Metrics and clearer logs for denied scaling

[sharnoff]

Problem description / Motivation

We don't have clear signals for when the scheduler denies upscaling, or the vm-monitor denies downscaling.

This came up here: https://neondb.slack.com/archives/C03TN5G758R/p1716167732610919?thread_ts=1716166712.233079

Feature idea(s) / DoD

The autoscaler-agent can:

1. Expose new instance-level metrics for amount of scaling requests that were not satisfied
2. Change the log level of scheduler responses that didn't grant the full request to "warn", rather than just "info"

Implementation ideas

Should be simple enough to inject these alongside the existing logs & metrics, wherever they're defined (IIRC exec_bridge / executor ?)