fix: Inconsistent behaviour with lifetime elision on tracked fn

[ShoyuVanilla]

Fixes #519

I think that there would be roughly two options for fixing this;

1. Returning error while proc macro expansion
2. Inserting a type alias like type __T<'db> = DbInputArgType; inside the expanded function's body so that the expanded code fails to compile

In this PR, I chose the first one because the second is little bit non-intuitive and the expanded code is slightly more bloated.
But the second one also has its own merit; It can utilize rustc's more detailed error message and is more consistent with the case that the other input parameter has an elided lifetime.
So, I'm okay to switch to the second one if I have to 😄

[netlify]

✅ Deploy Preview for salsa-rs canceled.

Name	Link
🔨 Latest commit	57f38aa
🔍 Latest deploy log	https://app.netlify.com/sites/salsa-rs/deploys/670fe3a07841da0008accd0e

[codspeed-hq]

CodSpeed Performance Report

Merging #596 will not alter performance

_{Comparing ShoyuVanilla:issue-519 (57f38aa) with master (c6c51a0)}

Summary

✅ 8 untouched benchmarks

[nikomatsakis]

I like it.

[nikomatsakis]

Nit: one test