[Task]: Refactor the FIM tests for all providers to be more deterministic

[rdimitrov]

Description

This is motivated by the fact that we are running a simpler model for the integration tests (0.5B). This sometimes results in responses which are not what we expect in the integration tests for FIM and so it fails the tests.

Instead of asking it to generate functions or code, we should refactor the tests to something simpler, i.e. "hello..." -> "hello world"

Additional Context

No response

[rdimitrov]

With the recent changes I think we are stable enough for now. If this changes, we'll adjust accordingly 👍