fix: also use agent on first preflight request

[Roemer]

Closes #28423

We use a proxy in order to communicate with the "outside world". Now the preflight first tries a request to api-proxy.cypress.io but with the agent explicitly set to null. When this request fails, it retries with api.cypress.io and this time with the agent set.
I suspect that the agent has the proxy info and therefore, the first request does not use the proxy and depending on the local setup, might wait up to 1 minute before aborting and retrying with the proxy. This leads to +1 minute to every cypress run.

[cypress-app-bot]

• Create a Draft Pull Request if your PR is not ready for review. Mark the PR as Ready for Review when you're ready for a Cypress team member to review the PR.

[Roemer]

This is extremely limiting us, as we run Cypress for a lot of projects and all of them now have 1 minute additional time for each project and I did not find a way to decrease the timeout from SIXTY_SECONDS with config or environment variables.
As the only difference between the call that works (makeReq({ baseUrl: apiUrl, agent })) and the one that does not work (makeReq({ baseUrl: preflightBaseProxy, agent: null })) is the agent, I am fairly sure that this is the issue.
Is there anything else needed to have its PR checked?

[jennifer-shehane]

@Roemer Thanks for opening a PR. We'll need a test to confirm the behavior before merging and we'll take a look to review.

[cypress-app-bot]

This PR has not had any activity in 180 days. If no activity is detected in the next 14 days, this PR will be closed.

[Roemer]

This PR is still relevant.

[jennifer-shehane]

@Roemer Oh boy, this definitely got lost. Sorry. We'll need a test to verify this behavior in order to accept the PR. Could you write a test to verify and prevent regressions?

[Roemer]

Pity the PR is closed as the bug still persists. As mentioned already, it is pretty much impossible to create a test as this is dependent on a specific (corporate) network setup. Having a networking-aware developer look at it should make it fairly obvious that it is currently a bug.