Add an eval binary that evaluates our semantic index against CodeSearchNet (#17375)

This PR is the beginning of an evaluation framework for our AI features.
Right now, we're evaluating our semantic search feature against the
[CodeSearchNet](https://github.com/github/CodeSearchNet) code search
dataset. This dataset is very limited (for the most part, only 1 known
good search result per repo) but it has surfaced some problems with our
search already.

Release Notes:

- N/A

---------

Co-authored-by: Jason <jason@zed.dev>
Co-authored-by: Jason Mancuso <7891333+jvmncs@users.noreply.github.com>
Co-authored-by: Nathan <nathan@zed.dev>
Co-authored-by: Richard <richard@zed.dev>

crates/evals/build.rs

@@ -0,0 +1,14 @@
+fn main() {
+    if cfg!(target_os = "macos") {
+        println!("cargo:rustc-env=MACOSX_DEPLOYMENT_TARGET=10.15.7");
+
+        println!("cargo:rerun-if-env-changed=ZED_BUNDLE");
+        if std::env::var("ZED_BUNDLE").ok().as_deref() == Some("true") {
+            // Find WebRTC.framework in the Frameworks folder when running as part of an application bundle.
+            println!("cargo:rustc-link-arg=-Wl,-rpath,@executable_path/../Frameworks");
+        } else {
+            // Find WebRTC.framework as a sibling of the executable when running outside of an application bundle.
+            println!("cargo:rustc-link-arg=-Wl,-rpath,@executable_path");
+        }
+    }
+}