Fixes #514 by checking if NO_PROXY contains DD_SITE domain, and the the user is likely relying on privatelink endpoints + private DNS isntead of an HTTP proxy for telemetry data.

This is kind of a hack in 2 ways:

1. reqwest has some support for NO_PROXY but hyper-proxy doesn't. I could use the built in version, but either way I'd need to build it myself
2. We simply pass a proxy string to libdatadog for APM + statsd, we don't actually pass config anywhere. So we'd have to add that.

This solution works for the customer reporting it, as their NO_PROXY string includes datadoghq.com.

But I wonder if there's a use case for a customer who would use privatelink for, say metrics only - but then use an HTTP proxy for logs (perhaps to inspect logs for PII or some other form of compliance).

In that case our entire shared http client architecture (complete with shared threadpools) would need to change. I think we should leave that for the future at this time, and help our friends in #514 move on.

Screenshots of my config:

VPC endpoints as per our docs:

Logs/metrics/traces: