fix: raise geolocation API timeout and retry the primary API once

[NoiceHax]

Description

The geolocation lookup used a hardcoded 0.5 second timeout for the primary API (get.geojs.io) and for the backup one (ipinfo.io). Half a second is a small budget for a network call, so a healthy API can miss it and codecarbon drops to the backup, or to the hardcoded Canada default when the backup is slow too.

The timeout now lives in a GEO_API_TIMEOUT constant set to 5 seconds, following the same pattern as ELECTRICITYMAPS_API_TIMEOUT in codecarbon/core/electricitymaps_api.py. The primary API also gets one retry, limited to timeouts and connection errors, so a reply that arrives but does not parse still goes to the backup like before. The Canada default is untouched, its warning now says plainly that emissions will be computed with the Canadian carbon intensity.

Related Issue

Fixes #854

Motivation and Context

The reporter started eight trackers in parallel and six of them logged "Unable to access geographical location through primary API". Sleeping up to 10 seconds between processes was the workaround. With half a second to answer, a handful of trackers hitting the same API at once is enough to miss the deadline, and the wrong country means the wrong carbon intensity for the whole run.

How Has This Been Tested?

I ran the package suite on Windows with CODECARBON_ALLOW_MULTIPLE_RUNS=True pytest -m "not integ_test" tests/, and 614 of the 615 collected tests pass. The one failure, test_task_energy_with_live_update_interference, fails the same way on master on this machine: it mocks CPU and RAM but not the GPU, and this laptop has a real NVIDIA card whose energy lands in the total. Three test modules need optional dev packages I could not install here, so they were left out of the run.

I added two tests in tests/test_geography.py and checked that both fail before the change. One has the primary API time out on the first call and answer on the second, asserting the backup is never called. The other asserts the request goes out with the constant instead of 0.5. I also bumped the expected request count in test_carbon_tracker_timeout, which counted the calls made when both APIs time out. black and isort are clean on the changed files.

Screenshots (if appropriate):

Not applicable.

Types of changes

What types of changes does your code introduce? Put an x in all the boxes that apply:

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

AI Usage Disclosure

Please refer to docs/how-to/ai-policy.md for detailed guidelines on how to disclose AI usage in your PR. Accurately completing this section is mandatory.

• 🟥 AI-vibecoded: You cannot explain the logic. Car analogy : the car drive by itself, you are outside it and just tell it where to go.
• 🟠 AI-generated: Car analogy : the car drive by itself, you are inside and give instructions.
• â­� AI-assisted. Car analogy : you drive the car, AI help you find your way.
• â™»ï¸� No AI used. Car analogy : you drive the car.

An AI coding agent wrote this patch and this description under my direction, and I read every line, ran the tests myself and can explain the change.

Checklist:

Go over all the following points, and put an x in all the boxes that apply.

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have read the docs/how-to/contributing.md document.
• I have added tests to cover my changes.
• All new and existing tests passed.