Emit HTTP timings

[patrobinson]

Description

When debugging network timeouts it can be useful to have detailed timings to know where specifically the time being consumed.

Changes

Add a TraceHTTP flag (only on agent start for now).
Use http trace to capture timings between each network event In the case of a network error emit this to error log. Otherwise emit it to the debug logs.

Example:

2024-09-17 09:31:11 DEBUG  HTTP Timing Trace uri=https://agent.buildkite.com/v3/ping method=GET hostPort=agent.buildkite.com:443 getConn=746.458µs dnsStart=773.625µs dnsDone=5.492208ms connectStart.tcp.3.168.86.46:443=5.502791ms connectDone.tcp.3.168.86.46:443=169.197333ms tlsHandshakeStart=169.207541ms tlsHandshakeDone=337.926125ms gotConn=337.964208ms reused=false idle=false idleTime=0s wroteHeaders=337.995208ms wroteRequest=337.995791ms gotFirstResponseByte=665.737916ms agent_version=3.81.0+x..dirty

Testing

• Tests have run locally (with go test ./...). Buildkite employees may check this if the pipeline has run automatically.
• Code is formatted (with go fmt ./...)

[wolfeidau]

@patrobinson are you going to include the status code? or even entire status?