Avoid duplicate Documents for same file

[varungandhi-src]

Originally flagged by @Strum355 in Slack:

It looks like we emit multiple SCIP Documents in the same Index for the same file, but with different relative paths. Out of these, some relative paths are wrong because they are not normalized.

We should make sure to normalize all relative paths and not do duplicate work for the same file.

Example URL

Scope

1. Normalize server side
2. No db migration

[varungandhi-src]

Hmm, I tried indexing the linked repo with the bad paths, and I cannot reproduce the issue:

git clone https://github.com/deepak2431/cody.git
cd cody && git checkout f32232e6aae6025ea2dc53af9288988c5b6a42eb
node ~/Code/scip-typescript/dist/src/main.js index --pnpm-workspaces --no-global-caches
~/Code/scip/scip print --json index.scip | jq '.documents[].relativePath' | grep '\.\.'

Would need to look deeper into where these bad paths are coming from.

[kritzcreek]

Another instance of this here: https://github.com/sourcegraph/sourcegraph/issues/61245

[varungandhi-src]

Something interesting:

• On S2, this works fine: https://sourcegraph.sourcegraph.com/github.com/sourcegraph/cody@032645f0e65c3d7958ff9d8fb0f8b78c56091062/-/blob/vscode/src/chat/chat-view/SimpleChatPanelProvider.ts?L622:19-622:34#tab=references
• On Sourcegraph.com, it doesn't quite work: https://sourcegraph.com/github.com/sourcegraph/cody@032645f0e65c3d7958ff9d8fb0f8b78c56091062/-/blob/vscode/src/chat/chat-view/SimpleChatPanelProvider.ts?L622:19-622:34#tab=references

Even though the same index should be getting uploaded to both places.

[mmanela]

I believe a minimal repo that will have this issue is shown in this sample I made. The key is having a child folder that uses the references node in tsconfig to refer to the parent folder.