Promote synthetic definitions into normal definitions if they share names

[olafurpg]

Fixes #403

Previously, we generated synthetic definitions for all synthetic
definitions. However, this meant that "find references" never showed
results for synthetic symbols share the same name as the non-synthetic
definition. This commit changes that so synthetic definitions that share
the same name as the non-synthetic symbol are included with "find
references" results.

While testing this change I noticed that we don't handle synthetic
constructor parameters so I added support for those along the way.

Test plan

Snapshot tests included in the diff. I also tested the change manually in this URL here https://sourcegraph.com/github.com/sourcegraph/sbt-sourcegraph@e2f71270ea518766c65097b1f17e97b55059d792/-/blob/example/foo.scala?L3:13#tab=references

[olafurpg]

@bobheadxi The pr-auditor check failed with a cryptic error message https://github.com/sourcegraph/lsif-java/runs/5334788397?check_suite_focus=true#step:4:34

![CleanShot 2022-02-25 at 15 20 51@2x](https://user-images.githubusercontent.com/1408093/155731053-74904d20-db13-4e49-b25a-2ebe55f1f41f.png) Error:<nil>}
[35](https://github.com/sourcegraph/lsif-java/runs/5334788397?check_suite_focus=true#step:4:35)
2022/02/25 14:23:17 preMergeAudit: CreateStatus: POST https://api.github.com/repos/sourcegraph/lsif-java/statuses/0c51ae30ffd4e50b1f5fd0a2dbf47e039a3c6b09: 401 Bad credentials []

Any advice on how to troubleshoot that?

[bobheadxi]

@olafurpg we use sourcegraph-bot to make API requests, does the bot have access to this repo?

[bobheadxi]

Ah I see - @olafurpg this PR seems to be from a fork, I'm not sure how API access to set status updates on a fork's commits works

[wchau]

In https://sourcegraph.com/github.com/sourcegraph/sbt-sourcegraph@e2f71270ea518766c65097b1f17e97b55059d792/-/blob/example/foo.scala?L3:13#tab=references

When I hover

B

in

case class B

It sometimes gives case class B... and sometimes gives object B.

[olafurpg]

@wchau I'm able to reproduce! That's an issue outside of lsif-java that needs to be fixed in the Sourcegraph backend. Both the case class and the object are defined at the same location and there appears to b non-determinism related to which symbol docstring gets displayed in the hover. I've added a reminder to report an upstream issue in the sourcegraph/sourcegraph erpo

[olafurpg]

Ah I see - @olafurpg this PR seems to be from a fork, I'm not sure how API access to set status updates on a fork's commits works

I can open PRs from the main repo but what should the pr-auditor do for PRs from external contributors? I have a preference for opening PRs from my fork instead of the main repo because I like push a lot of WIP branches while keeping the main repo clean. I don't like cloning a repo with lots of WIP branches from the maintainers.

[olafurpg]

@wchau I opened a followup issue to address the non-deterministic hovers https://github.com/sourcegraph/sourcegraph/issues/32044