Cap concurrency in location_to_hg_changeset_id

Summary:
⚠️ **This diff was generated by M-O** -- an AI code-quality bot that
produces one small, surgical Rust cleanup per day in fbcode/eden/mononoke.
walle_clean

location_to_hg_changeset_id uses future::try_join_all to convert N bonsai
changeset IDs to Hg IDs after resolving a commit-graph location. try_join_all
polls all N futures concurrently with no bound.

The count parameter comes from the client's commit_location_to_hash request
and can represent hundreds of changesets (Sapling's segment-based protocol
sends linear segments of 100-500+ entries). Unbounded fan-out on user-controlled
N risks saturating the derived-data SQL store and is flagged as a production
anti-pattern (S493741, S552025).

Replace try_join_all with stream::iter(...).buffered(100) to cap in-flight
derivation calls at 100. buffered (not buffer_unordered) preserves the
ordering of the result vec, matching the original semantics. Remove the now-
unused futures::future import.

Differential Revision: D103610505

fbshipit-source-id: 08e52faed4af38c45157277d55589cc3758d0f1c

Author: lmvasquezg

eden/mononoke/mononoke_api_hg/src/repo.rs

@@ -39,7 +39,6 @@ use futures::StreamExt;
 use futures::TryStream;
 use futures::TryStreamExt;
 use futures::compat::Stream01CompatExt;
-use futures::future;
 use futures::stream;
 use futures_util::try_join;
 use hgproto::GettreepackArgs;
@@ -530,12 +529,16 @@ impl<R: MononokeRepo> HgRepoContext<R> {
             .repo_ctx()
             .location_to_changeset_id(cs_location, count)
             .await?;
-        let hg_id_futures = result_csids
-            .iter()
-            .map(|result_csid| self.repo().derive_hg_changeset(self.ctx(), *result_csid));
-        future::try_join_all(hg_id_futures)
+        stream::iter(result_csids)
+            .map(|bcs_id| async move {
+                self.repo()
+                    .derive_hg_changeset(self.ctx(), bcs_id)
+                    .await
+                    .map_err(MononokeError::from)
+            })
+            .buffered(100)
+            .try_collect()
             .await
-            .map_err(MononokeError::from)
     }
 
     /// This provides the same functionality as