Run Performance bench on more platforms (#8312)

## Problem

https://github.com/neondatabase/cloud/issues/14721

## Summary of changes

add one more platform to benchmarking job 


57535c039c/.github/workflows/benchmarking.yml (L57C3-L126)

Run with pg 16, provisioner k8-neonvm by default on the new platform.

Adjust some test cases to

- not depend on database client <-> database server latency by pushing
loops into server side pl/pgSQL functions
- increase statement and test timeouts

First successful run of these job steps 

https://github.com/neondatabase/neon/actions/runs/9869817756/job/27254280428

test_runner/performance/test_hot_page.py

@@ -16,20 +16,34 @@ from pytest_lazyfixture import lazy_fixture
 )
 def test_hot_page(env: PgCompare):
     # Update the same page many times, then measure read performance
-    num_writes = 1000000
 
     with closing(env.pg.connect()) as conn:
         with conn.cursor() as cur:
             cur.execute("drop table if exists t, f;")
+            num_writes = 1000000
 
-            # Write many updates to the same row
+            # Use a PL/pgSQL block to perform many updates to the same row
+            # without depending on the latency between database client and postgres
+            # server
+            # - however a single staement should not run into a timeout so we increase it
+            cur.execute("SET statement_timeout = '4h';")
             with env.record_duration("write"):
-                cur.execute("create table t (i integer);")
-                cur.execute("insert into t values (0);")
-                for i in range(num_writes):
-                    cur.execute(f"update t set i = {i};")
+                cur.execute(
+                    f"""
+                DO $$
+                BEGIN
+                    create table t (i integer);
+                    insert into t values (0);
 
-            # Write 3-4 MB to evict t from compute cache
+                    FOR j IN 1..{num_writes} LOOP
+                        update t set i = j;
+                    END LOOP;
+                END $$;
+                """
+                )
+
+            # Write ca 350 MB to evict t from compute shared buffers (128 MB)
+            # however it will still be in LFC, so I do not really understand the point of this test
             cur.execute("create table f (i integer);")
             cur.execute("insert into f values (generate_series(1,100000));")
 

test_runner/performance/test_hot_table.py

@@ -16,8 +16,8 @@ from pytest_lazyfixture import lazy_fixture
 )
 def test_hot_table(env: PgCompare):
     # Update a small table many times, then measure read performance
-    num_rows = 100000  # Slightly larger than shared buffers size  TODO validate
-    num_writes = 1000000
+    num_rows = 100000  # initial table size only about 4 MB
+    num_writes = 10000000  # write approximately 349 MB blocks > 128 MB shared_buffers
     num_reads = 10
 
     with closing(env.pg.connect()) as conn:
@@ -28,8 +28,21 @@ def test_hot_table(env: PgCompare):
             with env.record_duration("write"):
                 cur.execute("create table t (i integer primary key);")
                 cur.execute(f"insert into t values (generate_series(1,{num_rows}));")
-                for i in range(num_writes):
-                    cur.execute(f"update t set i = {i + num_rows} WHERE i = {i};")
+                # PL/pgSQL block to perform updates (and avoid latency between client and server)
+                # - however a single staement should not run into a timeout so we increase it
+                cur.execute("SET statement_timeout = '4h';")
+                cur.execute(
+                    f"""
+                DO $$
+                DECLARE
+                    r integer := {num_rows};
+                BEGIN
+                    FOR j IN 1..{num_writes} LOOP
+                        UPDATE t SET i = j + r WHERE i = j;
+                    END LOOP;
+                END $$;
+                """
+                )
 
             # Read the table
             with env.record_duration("read"):