[SYCL][E2E] Fix sycl-ls discovery

[npmiller]

We should also look for sycl-ls in the PATH and not just in the llvm tools dir.

[github-actions]

✅ With the latest revision this PR passed the Python code formatter.

[aelovikov-intel]

Why is that so? Are you trying to use sycl-ls outside of the build?

[aelovikov-intel]

+ @frasercrmck

[npmiller]

Why is that so? Are you trying to use sycl-ls outside of the build?

Specifically, I'm running E2E tests standalone against the release package's clang++, it's located in bin/compiler rather than bin, so the current lit setup isn't able to find sycl-ls.

But more broadly this seems to be the only restrictions to using a compiler and a runtime installed in different locations when running the tests.

[aelovikov-intel]

We have some tight coupling between sycl-ls output format and the E2E infrastructure, so I wouldn't want to pick some random version of it from the PATH...

Can we make it more explicit for the user, maybe by asking to provide the path to sycl-ls explicitly? WDYT?

[npmiller]

Closing this.

We've ended up deciding it didn't make that much sense to run tests against the clang++ binary from the release package so we don't need this anymore.

If we do end up revisiting it at some point I do agree that a flag to point to a specific sycl-ls binary would make sense.