Clean Staking <-> Session tests and Use a Realistic Scenario

[kianenigma]

Currently, almost all staking tests use this function

substrate/frame/staking/src/mock.rs

Lines 671 to 679 in 918e58e

	pub fn start_session(session_index: SessionIndex) {
	assert_eq!(<Period as Get<BlockNumber>>::get(), 1, "start_session can only be used with session length 1.");
	for i in Session::current_index()..session_index {
	Staking::on_finalize(System::block_number());
	System::set_block_number((i + 1).into());
	Timestamp::set_timestamp(System::block_number() * 1000);
	Session::on_initialize(System::block_number());
	Staking::on_initialize(System::block_number());
	}

To which I have added the assertion that demonstrates the issues; Up until this commit, staking's mock didn't even have a configuration to set the session length and it was always 1. The assert should prevent us from creating a bug here, but ideally the assumption should be removed and tests that use it should be revised. There should be tests that check staking's session rotation with session/era lengths of N and M, where both of them are greater than 1.

[kianenigma]

Also, the first era contains session 0, which has length 0. A test should properly demonstrate this.

[kianenigma]

long term i'd really like to see a revamp of staking tests so that:
1- active_era and current_era are actually different.
2- sessions is actually queuing the receiving validators (same as all of our real chains).