Set CMAKE_BUILD_PARALLEL_LEVEL to num_cpus

cmake-rs' jobserver doesn't work reliably, or at all. One workaround is
to set CMAKE_BUILD_PARALLEL_LEVEL to num_cpus. On my machine it makes
builds over 3x faster

Cargo.toml

@@ -42,3 +42,4 @@ linked_hash_set = "0.1"
 once_cell = "1.0"
 tower = "0.4"
 tower-layer = "0.3"
+num_cpus = "1.16.0"

boring-sys/Cargo.toml

@@ -81,3 +81,4 @@ bindgen = { workspace = true }
 cmake = { workspace = true }
 fs_extra = { workspace = true }
 fslock = { workspace = true }
+num_cpus = { workspace = true }

boring-sys/build/main.rs

@@ -554,6 +554,8 @@ fn built_boring_source_path(config: &Config) -> &PathBuf {
 
         let mut cfg = get_boringssl_cmake_config(config);
 
+        cfg.env("CMAKE_BUILD_PARALLEL_LEVEL", num_cpus::get().to_string());
+
         if config.features.fips {
             let (clang, clangxx) = verify_fips_clang_version();
             cfg.define("CMAKE_C_COMPILER", clang)