Drop benchmarking

I don't really look at the results of the benchmarks anyway, so having
them in the repo creates a false sense of benchmarkdness.

If I get to implementing proper benchmarking, I'd probably stay away
from criterion -- we need something much much simpler for this crate.

Author: matklad

src/tools/rust-analyzer/lib/smol_str/Cargo.toml

@@ -14,8 +14,3 @@ serde = { version = "1", optional = true, default_features = false, features = [
 proptest = "0.10"
 serde_json = "1"
 serde =  { version = "1", features = [ "derive" ] }
-criterion = "0.2"
-
-[[bench]]
-name = "building"
-harness = false