tests/flaky: ChainHead operation limits test is flaky

[lexnv]

The ensure_operation_limits_works is flaky:

• Step1: Subscribe to chainHead_v1_follow with maximum of 1 ongoing operations per subscription
• Step2: Make a storage query via chainHead_v1_storage
• Step3: Await OperationStorageDone and by this point, the capacity should be released
• Step4: Make a runtime API call

polkadot-sdk/substrate/client/rpc-spec-v2/src/chain_head/tests.rs

Line 2898 in 31abe61

MethodResponse::LimitReached => panic!("Expected started response"),

It looks like the capacity reserved by chainHead_v1_storage is not released by the time make a runtime API call:

--- TRY 3 STDERR:        sc-rpc-spec-v2 chain_head::tests::ensure_operation_limits_works ---
thread 'chain_head::tests::ensure_operation_limits_works' panicked at substrate/client/rpc-spec-v2/src/chain_head/tests.rs:2898:41:
Expected started response
stack backtrace:
   0: rust_begin_unwind
   1: core::panicking::panic_fmt
   2: sc_rpc_spec_v2::chain_head::tests::ensure_operation_limits_works::{{closure}}
   3: tokio::runtime::scheduler::current_thread::Context::enter
   4: tokio::runtime::context::scoped::Scoped<T>::set
   5: tokio::runtime::scheduler::current_thread::CoreGuard::block_on
   6: tokio::runtime::context::runtime::enter_runtime
   7: tokio::runtime::runtime::Runtime::block_on
   8: sc_rpc_spec_v2::chain_head::tests::ensure_operation_limits_works
   9: core::ops::function::FnOnce::call_once
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

Detected by: https://github.com/paritytech/polkadot-sdk/actions/runs/13154495481/job/36708551235