bench_local IDs must be refs in rust-lang/rustc

To reproduce:
./target/release/collector bench_local $(rustup +nightly which rustc) testrun

Expected:
The run ID "testrun" should label the current run in the database

Actual:
It's also used as a ref when checking out the rustc repo and the whole collector run stops on this benchmark:

15 benchmarks remaining
Cloning into 'rust'...
remote: Enumerating objects: 1406657, done.
remote: Total 1406657 (delta 0), reused 0 (delta 0), pack-reused 1406657
Receiving objects: 100% (1406657/1406657), 580.28 MiB | 2.47 MiB/s, done.
Resolving deltas: 100% (1133476/1133476), done.
fatal: ambiguous argument 'testrun': unknown revision or path not in the working tree.
Use '--' to separate paths from revisions, like this:
'git <command> [<revision>...] -- [<file>...]'
thread 'main' panicked at 'git reset --hard successful', collector/src/execute/rustc.rs:50:5

Probably caused by these commits: https://github.com/rust-lang/rustc-perf/commit/965ad78b70514a29098ad57fb33601873716ca67
https://github.com/rust-lang/rustc-perf/commit/9eef993b9487784c79792573d33c40bcdf98528b (when the repo already exists locally)

I'd suggest either having bench_local runs default to checking out origin/master, or adding a new flag to collector to represent a ref in the rust-lang repo (bit more involved)

If someone can verify this is an issue then I can submit a fix for it.

Reproduce the ./target/release/collector bench_local command from the issue, then inspect collector/src/execute/rustc.rs around line 50 and the referenced commits. Verify that the run ID still labels the database run while rustc checkout uses a valid repository ref, and confirm the benchmark run completes.