fix(auth): use the timeout-bounded HTTP client for workspace fetch and auth probe

Follow-up from #230: cache_workspaces, api_key_authorized_workspaces,
and credentials::check_status used bare reqwest::blocking::Client::new(),
which has no request timeout — a stalled server could hang 'auth login'
or the 4xx auth-status probe (which runs inside error printing)
indefinitely. They now share raw_http::build_http_client() (300s cap +
TCP keepalive) like every other raw-HTTP path. Test-only localhost
clients are left as-is.

Author: eddietejeda

src/client/credentials.rs

@@ -39,7 +39,7 @@ pub fn check_status(profile_config: &config::ProfileConfig) -> AuthStatus {
         };
 
     let url = format!("{}/workspaces", profile_config.api_url);
-    let client = reqwest::blocking::Client::new();
+    let client = crate::client::raw_http::build_http_client();
     let req = client
         .get(&url)
         .header("Authorization", format!("Bearer {access_token}"));

src/commands/auth.rs

@@ -440,7 +440,7 @@ fn cache_workspaces(
     access_token: &str,
 ) -> Result<Vec<config::WorkspaceEntry>, String> {
     let url = format!("{}/workspaces", profile.api_url);
-    let client = reqwest::blocking::Client::new();
+    let client = crate::client::raw_http::build_http_client();
     let req = client
         .get(&url)
         .header("Authorization", format!("Bearer {access_token}"));
@@ -485,7 +485,7 @@ fn api_key_authorized_workspaces(
         return Vec::new();
     };
     let url = format!("{}/workspaces", profile_config.api_url);
-    let client = reqwest::blocking::Client::new();
+    let client = crate::client::raw_http::build_http_client();
     let req = client
         .get(&url)
         .header("Authorization", format!("Bearer {access_token}"));