improve/client-buff-alloc-fix

[JKearnsl]

optimize heap

[JKearnsl]

The following test demonstrates the problem well

#[test]
fn buffer_reuse() {
    let pool = BufferPool::new(Some(214748));
    let mut results = vec![(0u128, 0u128); 214748];
    let mut buffers= Vec::new();
    for i in 0..214748 {
        let mut buf = pool.alloc().unwrap();
        buf.write(&[i as u8; 9000]).unwrap();
        buffers.push(buf);
    }
    for i in buffers.into_iter() {
        pool.release(i);
    }
    
    let mut start = std::time::Instant::now();
    for i in 0..214748 {
        start = std::time::Instant::now();
        let mut buf = pool.alloc().unwrap();
        results[i] = (results[i].0, start.elapsed().as_nanos());
        pool.release(buf);
    }

    for i in 0..214748 {
        start = std::time::Instant::now();
        let buf = Vec::<u8>::with_capacity(9000);
        results[i] = (start.elapsed().as_nanos(), results[i].1);
    }
    for (alloc_time, reuse_time) in results {
        if alloc_time > 1000 || reuse_time > 1000 {
            panic!("Allocation or reuse took too long: alloc={}ns, reuse={}ns", alloc_time, reuse_time);
        }
    }
}

Panic:

warning: `client` (lib test) generated 3 warnings (run `cargo fix --lib -p client --tests` to apply 1 suggestion)
    Finished `release` profile [optimized] target(s) in 0.08s
     Running unittests src/lib.rs (target/release/deps/client-ffaab629ddbc9824)

Allocation or reuse took too long: alloc=27814ns, reuse=30ns
thread 'runtime::buffer::tests::buffer_reuse' panicked at crates/client/src/runtime/buffer.rs:152:17:
Allocation or reuse took too long: alloc=27814ns, reuse=30ns

In the first iterations, vector reuse may seem slow, but in subsequent iterations, the reuse-based implementation wins.

alloc=27814ns vs reuse=30ns