fix a bug about snapshot and adjust some test cases

[hicqu]

The bug causes an exmaple panics in #205 .

In our implemetation, RaftLog::term(un_received_index) returns 0. So if a peer receives a snapshot or normal entry with log-term 0, RaftLog::match_term(log_term, un_received_index) will return true but it should be false exactly.

This PR fixes it by making RaftLog::match_term cover this case.

And, add a case test_snapshot_with_term_0 to cover it, and do some adjusts for some other test cases. At last, remove Interface::initial because we don't need it any more.

[hicqu]

PTAL @Hoverbear @BusyJay thank you!

[Hoverbear]

Suggested change

	// Apply the snashot. It's also necessary with same reason as above.
	// Apply the snapshot. It's also necessary with same reason as above.

Also, which reason? That the program will be blocked forever?

[Hoverbear]

Suggested change

	pub fn base_config() -> Config {
	pub fn default_config() -> Config {

[Hoverbear]

@hicqu Seems maybe RaftLog::term(un_received_index) should return Option<u64> instead? Would that be a good way to fix this problem?

[hicqu]

Thanks for your review @Hoverbear ! I will address them in #214 . Now I will close this PR because I prefer the another solution. :)

[bytewatch]

I think cause the root cause of the panic of five_mem_node is MemStorage::initialize_with_conf_state() :

        // To fix this panic,  add a more line: `core.raft_state.hard_state.set_commit(1)` , 
        // and handle the case that `prs.get_mut(self.id)` is None in restore_snapshot.
        // Or , just commet these two lines: 
        core.snapshot_metadata.set_index(1);  
        core.raft_state.hard_state.set_commit(1);  
        core.raft_state.conf_state = ConfState::from(conf_state);

[bytewatch]

I don't think it need to add some logics that exceeds etcd-raft.