Install wrapper script for Go on Linux to support tracing for Go 1.21 and above

[mbg]

This PR modifies the init Action to add a go wrapper script to the PATH if we are running on Linux in order to better cope with tracing builds for Go 1.21 and above. This works as follows:

• If we are initialising CodeQL for Go and we are running on Linux, then we obtain the current path to the go binary.
• We query the type of executable with the file command.
• If this contains the string "statically linked", we install the wrapper script to a bin directory in the CodeQL temporary directory.
• We add that bin directory to the system PATH
• We store the location of the wrapper script in a new environment variable, CODEQL_ACTION_GO_BINARY (we could probably omit this since the path can be reconstructed in the analyze step)

Additionally, we verify that the result of which go points to our wrapper script in the analyze step by checking it against the value of CODEQL_ACTION_GO_BINARY to make sure that users did not e.g. use a setup-go step after the init action.

Merge / deployment checklist

• Confirm this change is backwards compatible with existing workflows.
• Confirm the readme has been updated if necessary.
• Confirm the changelog has been updated if necessary.

[criemen]

Can we already add a feature gate/version flag to that, so that we automatically stop this wrapper script installation at some point in the future? Or a turn-off switch from the CLI?
Once we fix this in the CLI, we need to be able to retrospectively turn off this hack from the CLI, otherwise we'll extract go files with older codeql-action versions double.

Edit: Maybe we can use the new version feature already?

[smowton]

This broadly looks good but still needs some sort of version gating mechanism so that we don't start double-tracing if and when the CLI starts handling this situation for itself.

[henrymercer]

As a general comment, are we happy with the test coverage here? Specifically, do we want to add some unit tests or some specific assertions to the PR checks? Or do the existing checks passing on Go 1.21.1 give us enough confidence?

[henrymercer]

Thanks, I'd like to hear what Chris thinks about how we should behave in the absence of file, but otherwise this looks good to me.

[henrymercer]

Very nice!