feat: Extend outbox processor with operational metrics (`pulse.outbox.*`)

[samtrion]

User Story

As a platform engineer operating the outbox processor, I want detailed metrics for the outbox pipeline exposed via System.Diagnostics.Metrics, so that I can monitor queue depth, failure rates, and processing latency in production using OpenTelemetry-compatible tooling.

---

Background

The current ActivityAndMetricsRequestInterceptor covers mediator-level metrics (pulse.requests.total, pulse.request.duration, etc.), but the outbox processor background service emits no metrics. This creates a blind spot for operations teams monitoring message delivery health.

---

Requirements

• Extend the existing Meter ("NetEvolve.Pulse") in Defaults.cs with the following new instruments:

  Metric Name	Instrument Type	Unit	Description
  pulse.outbox.pending	ObservableGauge<long>	messages	Current number of pending outbox messages
  pulse.outbox.processed.total	Counter<long>	messages	Cumulative number of successfully processed messages
  pulse.outbox.failed.total	Counter<long>	messages	Cumulative number of failed processing attempts
  pulse.outbox.deadletter.total	Counter<long>	messages	Cumulative number of messages moved to dead-letter
  pulse.outbox.processing.duration	Histogram<double>	ms	Duration of each outbox processing batch

• OutboxProcessorHostedService must record all metrics during its polling loop.
• The pulse.outbox.pending gauge must query the IOutboxRepository for the current count.
• Metric recording must not throw; exceptions must be caught and logged at Warning level.

---

Acceptance Criteria

• All five metrics are registered and visible when the outbox processor is configured.
• pulse.outbox.pending reflects the actual value returned by IOutboxRepository.
• pulse.outbox.processed.total increments by the number of messages successfully sent in a batch.
• pulse.outbox.failed.total increments when a message transitions to Failed.
• pulse.outbox.deadletter.total increments when a message transitions to DeadLetter.
• pulse.outbox.processing.duration records the elapsed time per polling cycle as a histogram.
• Unit tests verify metric instrument registration and correct increment/observation behavior.
• Integration tests verify that metrics are non-zero after a processing cycle with messages.
• XML documentation is updated for any modified public or internal members.

---

Out of Scope

• Custom OpenTelemetry exporters.
• Dashboard or alerting configuration.
• Per-event-type metric dimensions/tags (may be addressed in a follow-up).