fix: adjust pool usage in tests and examples

sqlx-bench/README.md

@@ -15,7 +15,7 @@ This Cargo project implements various benchmarks for SQLx using
         the pool, with or without the pool being fair. Concurrently to the benchmark iteration
         function calling and blocking on `Pool::acquire()`, a varying number of background tasks are
         also calling `acquire()` and holding the acquired connection for 500µs each before releasing
-        it back to the pool. The pool is created with `.min_size(50).max_size(50)` so we shouldn't
+        it back to the pool. The pool is created with `.min_connections(50).max_connections(50)` so we shouldn't
         be measuring anything but the actual overhead of `Pool`'s bookeeping.
 
 ### Running

sqlx-bench/benches/pg_pool.rs

@@ -27,11 +27,11 @@ fn do_bench_acquire(b: &mut Bencher, concurrent: u32, fair: bool) {
             // we don't want timeouts because we want to see how the pool degrades
             .connect_timeout(Duration::from_secs(3600))
             // force the pool to start full
-            .min_size(50)
-            .max_size(50)
+            .min_connections(50)
+            .max_connections(50)
             // we're not benchmarking `ping()`
-            .test_on_acquire(false)
-            .fair(fair)
+            .test_before_acquire(false)
+            .__fair(fair)
             .build(
                 &dotenv::var("DATABASE_URL").expect("DATABASE_URL must be set to run benchmarks"),
             ),