token,scanner,parser,checker,cgen: add file_idx to token.Pos

[kbkpbot]

The original token.Pos does not include file name information.
This PR adds a file_idx field to token.Pos, allowing us to identify which source file each token belongs to.

[spytheman]

I've reverted it, because it caused a performance regression, without a clear benefit for correctness or usability:

[kbkpbot]

Occasionally, I encountered issues where error or warning messages pointed to incorrect code locations.

I believe this happens because the error function relies on c.file.path to work properly, but in some cases, this is not a reliable method for obtaining the correct filename.

I will make a note of these cases next time for testing purposes.

[kbkpbot]

I've reverted it, because it caused a performance regression, without a clear benefit for correctness or usability:

Can you share the script you use to compare performance?

[spytheman]

Can you share the script you use to compare performance?

The screenshot is from the #fast channel in the V discord server, but you can run something locally, that shows a similar result with:
./v run .github/workflows/compare_pr_to_master.v -prod

(executed on the branch that has the changes)

[spytheman]

I suspect, that the slowdown is caused mainly by the increased size of the Pos struct, since it is used a lot in practically every AST node.

If the new file index is encoded in the high 16 bits of the col field, the total size of ast.Pos will be the same, and perhaps the impact will be much lower.
Although I am not sure if just changing it to col u16 and idx i16 may not be enough. It will still allow for 65536 different columns, and 32767 different file indexes.