[chore] Source Timeout for Providers

[ankitpatel96]

Description:

Datadog exporter tests have an unnecessary 30 second wait while it waits for various metadata service calls to timeout (this is the default TCP timeout in the go network stack). Also had to modify the ec2 provider so that it respected the context. Most of the changes are just propagating this new timeout to the provider, and propagating this context around. I made the decision to differentiate the timeout context and the overall context for the chainProvider - these could be brought together but it looks like we had existing logic for a context timeout vs a chainProvider context cancellation so I wanted to respect that split.

Testing:
mostly stuck to a 31 second timeout on existing tests to not change behavior. Added a tighter timeout for some of the slower tests.
Added some tests for the timeout logic

[mx-psi]

Any reason to do this here and not at the top level (when we build the source provider)?

[ankitpatel96]

I'm a little confused - which part should we move where? Moving the creation of the child context into the Chain function?

[mx-psi]

Moving the WithTimeout up until we call Source instead of here. It does mean adding it in more places, which is a bit more annoying. I'll leave what to do up to you

Also, I am now wondering if we want a separate timeout for this or if just the per-OTLP blob timeout added by the exporterhelper works . Probably we want a separate one?

[ankitpatel96]

Ah - I think its good to leave in here. I think its good for this logic to be right next to the call site and if we ever want to do something fancier like per source timeout it will be much easier to add it here.

I definitely think we should want a separate timeout for this! the per otlp blob timeout should be used just for the normal export case - not the occasions we're blocked on fetching a source.