Marks local-cluster tests with #[serial]

[brooksprumo]

Problem

local-cluster tests timeout a lot; one is test_wait_for_max_stake(). All the local-cluster tests are marked with #[serial] to ensure they don't interfere with each other, but test_wait_for_max_stake() is not. We also see test failures in test_boot_from_local_state, which are due to timeouts too. Since #[serial] is opt-in, if we do not add the annotation, the test can run concurrently with other tests.

Summary of Changes

Mark all the local-cluster tests with #[serial].

[steviez]

@bw-solana - FYI, I know you've spent some time looking at test_wait_for_max_stake as well.
@brooksprumo - Some PR's where we've been looking at this test: #3295, #3430

[brooksprumo]

@steviez - Thanks for the links! Yeah, originally I was looking at this because my test, test_boot_from_local_state, times out sometimes too. I have code that bounds the timeouts to 30 seconds, so we don't wait forever, but it is still something that should never happen (we should only be waiting for about 10 slots, which should happen much faster than 30 seconds). I noticed that we don't have #[serial] on all the local-cluster tests, so there's possibility for issues. Low hanging fruit to ensure tests are not stepping on each other.

[bw-solana]

LGTM. Can monitor CI time to see if we need more shards

[steviez]

Better reliability / reduced flakiness would be a win and can always look at how to speed things back up after the fact