Improve instructions to agents for running tests (#23804)

## Summary

Several changes to our instructions that should, I think, make it easier
and faster for agents to run our tests:
- Suggest setting `CARGO_PROFILE_DEV_OPT_LEVEL=1` rather than
`--cargo-profile=fast-test`. `--cargo-profile` is only a valid command
if you're invoking the tests using `cargo nextest run`; if you're
invoking the tests using `cargo test`, you have to use
`--profile=fast-test` instead. That means that it sometimes takes three
tries for Claude to run our tests successfully if it's being invoked
from Claude Code for Web:
1. Try `cargo nextest run --cargo-profile=fast-test` <-- oh no, `cargo
nextest` isn't installed
2. Try `cargo test --cargo-profile=fast-test` <-- oh no,
`--cargo-profile` isn't a valid CLI argument for `cargo test`
  3. Try `cargo test --profile=fast-test` <-- hooray, it finally worked!
- Set two environment variables for `cargo-insta` so that snapshot tests
are easier for agents to handle. Together these are equivalent to
running `cargo insta test --accept`, but they should work even if
`cargo-insta` isn't installed (which it won't be in a Claude Code for
Web environment):
- `INSTA_FORCE_PASS` means that the test runner will keep going even if
it encounters a failing snapshot in an mdtest
- `INSTA_UPDATE=always` means that the snapshots are automatically
accepted, so the agent doesn't need to do a separate `cargo insta
review` step; it can just review the changes that have been applied to
the existing snapshot files.

## Test Plan

Merge it and see?

Author: AlexWaygood

AGENTS.md

@@ -4,41 +4,31 @@ This repository contains both Ruff (a Python linter and formatter) and ty (a Pyt
 
 ## Running Tests
 
-Run all tests (using `nextest` for faster execution):
+Run all tests (using `nextest` for faster execution, setting `CARGO_PROFILE_DEV_OPT_LEVEL=1` to enable optimizations while retaining debug info, and setting `INSTA_FORCE_PASS=1 INSTA_UPDATE=always` to ensure all snapshots are updated):
 
 ```sh
-cargo nextest run
-```
-
-For faster test execution, use the `fast-test` profile which enables optimizations while retaining debug info:
-
-```sh
-cargo nextest run --cargo-profile fast-test
+CARGO_PROFILE_DEV_OPT_LEVEL=1 INSTA_FORCE_PASS=1 INSTA_UPDATE=always cargo nextest run
 ```
 
 Run tests for a specific crate:
 
 ```sh
-cargo nextest run -p ty_python_semantic
+CARGO_PROFILE_DEV_OPT_LEVEL=1 INSTA_FORCE_PASS=1 INSTA_UPDATE=always cargo nextest run -p ty_python_semantic
 ```
 
-Run a single mdtest file:
+Run a single mdtest file. The path to the mdtest file should be relative to the `crates/ty_python_semantic/resources/mdtest` folder:
 
 ```sh
-cargo nextest run -p ty_python_semantic --test mdtest -- mdtest::<path/to/mdtest_file.md>
+CARGO_PROFILE_DEV_OPT_LEVEL=1 INSTA_FORCE_PASS=1 INSTA_UPDATE=always cargo nextest run -p ty_python_semantic -- mdtest::<path/to/mdtest_file.md>
 ```
 
 To run a specific mdtest within a file, use a substring of the Markdown header text as `MDTEST_TEST_FILTER`. Only use this if it's necessary to isolate a single test case:
 
 ```sh
-MDTEST_TEST_FILTER="<filter>" cargo nextest run -p ty_python_semantic --test mdtest -- mdtest::<path/to/mdtest_file.md>
+MDTEST_TEST_FILTER="<filter>" CARGO_PROFILE_DEV_OPT_LEVEL=1 INSTA_FORCE_PASS=1 INSTA_UPDATE=always cargo nextest run -p ty_python_semantic -- mdtest::<path/to/mdtest_file.md>
 ```
 
-Update snapshots after running tests:
-
-```sh
-cargo insta accept
-```
+After running the tests, always review the contents of any snapshots that have been added or updated.
 
 ## Running Clippy