Fixes#16484: rl fix failed tests

[gpoblon]

Fixed failing tests, and cleaned integration tests so compile.rs stays clean of code logic. Added tests comments to highlight that some tests could be meant to fail : uncomment to ensure tests are still working the way they are intended to.

https://issues.rudder.io/issues/16575

[amousset]

Added tests comments to highlight that some tests could be meant to fail : uncomment to ensure tests are still working the way they are intended to.

Should these tests always fail or is it a temporary situation? We should generally not require manual modification of the code to run all tests. To ensure proper failure modes, the test logic should be reversed, and return a success if the tested code fails as expected (a bit like the #[should_panic] annotation that allows ensuring a code path panics in a test).

[gpoblon]

These tests were mainly here to show you how tests are handled :
There are 2 different kinds of errors in the test :

• Result<_, Error> from the compile function. These are handled just as you say: if a file is prefixed by f_ and rudderc::compile() returns an error, then the test assertion will be Ok. Not Error.
• Errors from the test feature itself, ie: something went wrong doing a test (name of the test file, error when writing the file etc) so it HAS to fail in order to notify the user something is wrong with tests themselves.
  The commented tests fall into the latter kind. If you try to test a file name that has a bad format (ie not prefixed by s_ or f_) the test itself is supposed to fail.
  So they have no vocation to be uncommented unless you want to be sure that the integration test feature works as intended.

[Normation-Quality-Assistant]

OK, merging this PR