Add structured logging and CloudWatch EMF metrics

[nandanadileep]

Closes #6

Summary

• Every Lambda invocation now emits two log lines: a plain structured JSON entry (queryable in Logs Insights) and a CloudWatch Embedded Metrics Format line (auto-extracted as custom metrics — no API call, no added latency)
• Metrics published under the AlexaLLM namespace with Intent as the dimension: Invocations, Errors, Latency
• cloudwatch_dashboard.json provides a ready-to-import dashboard with 4 widgets

Files changed

File	Change
logger.py	log_invocation() + emit_metrics() using EMF
lambda_function.py	Handler timed; both logger functions called in finally block
cloudwatch_dashboard.json	4-widget dashboard: invocations by intent, avg latency, error rate %, P99 latency
test_lambda.py	TestLogger (13 cases) covering JSON structure, EMF format, error flag, and handler integration

One-time setup

Import the dashboard:

aws cloudwatch put-dashboard \
  --dashboard-name AlexaLLM \
  --dashboard-body file://cloudwatch_dashboard.json

Logs Insights query to see all invocations:

fields intent, userId, latency_ms, error
| filter event = "invocation"
| sort @timestamp desc

Test plan

• Invoke the skill — verify invocation JSON log appears in CloudWatch Logs
• Verify AlexaLLM namespace appears in CloudWatch Metrics within ~5 min
• Import dashboard JSON and confirm all 4 widgets populate
• Force an LLM error — verify Errors metric increments
• 192 unit tests pass