It could happen that i1 persists an extra entry before i2 gets it, and
if i2 starts requesting a vote over and over, it won't get that entry.

- Eliminate a couple of ancient TODOs. Extend the test.
- Fix `ConnectionPool::send` docstring appropriately.
- Tweak timeouts in test because it used to fail on my machine.

Also run `cargo update`

```
    Updating crates.io index
    Updating aho-corasick v0.7.18 -> v0.7.20
    Updating arc-swap v1.5.0 -> v1.5.1
    Updating assert_cmd v2.0.2 -> v2.0.7
    Updating autocfg v1.0.1 -> v1.1.0
    Updating base64 v0.13.0 -> v0.13.1
    Updating block-buffer v0.10.2 -> v0.10.3
    Updating bstr v0.2.17 -> v1.0.1
    Updating bytes v1.1.0 -> v1.3.0
    Updating cc v1.0.72 -> v1.0.77
    Removing chrono v0.4.19
    Updating clap v3.1.12 -> v3.2.23
    Updating clap_derive v3.1.7 -> v3.2.18
    Updating clap_lex v0.1.1 -> v0.2.4
    Updating cmake v0.1.48 -> v0.1.49
    Updating cpufeatures v0.2.1 -> v0.2.5
    Updating crossbeam-channel v0.5.1 -> v0.5.6
    Updating crossbeam-utils v0.8.5 -> v0.8.14
    Updating crypto-common v0.1.3 -> v0.1.6
    Updating ctor v0.1.21 -> v0.1.26
    Updating digest v0.10.3 -> v0.10.6
    Updating either v1.6.1 -> v1.8.0
      Adding fastrand v1.8.0
    Updating generic-array v0.14.4 -> v0.14.6
    Updating getrandom v0.2.3 -> v0.2.8
    Updating ghost v0.1.2 -> v0.1.6
    Updating hashbrown v0.11.2 -> v0.12.3
    Updating indexmap v1.8.1 -> v1.9.2
    Updating indoc v1.0.3 -> v1.0.7
      Adding instant v0.1.12
    Updating inventory v0.2.1 -> v0.2.3
    Updating itertools v0.10.3 -> v0.10.5
    Updating itoa v1.0.1 -> v1.0.4
    Updating libc v0.2.126 -> v0.2.138
    Updating log v0.4.14 -> v0.4.17
    Updating md-5 v0.10.1 -> v0.10.5
    Updating memchr v2.4.1 -> v2.5.0
    Removing nix v0.23.1
    Removing nix v0.24.1
      Adding nix v0.23.2
      Adding nix v0.24.3
    Removing num-integer v0.1.44
    Updating num-traits v0.2.14 -> v0.2.15
      Adding num_threads v0.1.6
    Updating once_cell v1.9.0 -> v1.16.0
    Updating os_str_bytes v6.0.0 -> v6.4.1
      Adding paste v1.0.10
    Updating ppv-lite86 v0.2.15 -> v0.2.17
    Updating predicates v2.1.0 -> v2.1.4
    Updating predicates-core v1.0.2 -> v1.0.5
    Updating predicates-tree v1.0.4 -> v1.0.7
    Updating proc-macro2 v1.0.43 -> v1.0.47
    Updating protobuf v2.27.1 -> v2.28.0
    Updating protobuf-codegen v2.27.1 -> v2.28.0
    Updating rand_core v0.6.3 -> v0.6.4
    Updating redox_syscall v0.2.10 -> v0.2.16
    Updating redox_users v0.4.0 -> v0.4.3
    Updating regex v1.5.4 -> v1.7.0
    Updating regex-syntax v0.6.25 -> v0.6.28
    Updating rmp v0.8.10 -> v0.8.11
    Updating rustversion v1.0.6 -> v1.0.9
    Updating ryu v1.0.9 -> v1.0.11
    Updating serde v1.0.133 -> v1.0.150
    Updating serde_bytes v0.11.5 -> v0.11.7
    Updating serde_derive v1.0.133 -> v1.0.150
    Updating serde_json v1.0.73 -> v1.0.89
    Updating slog-stdlog v4.1.0 -> v4.1.1
    Updating slog-term v2.8.0 -> v2.9.0
    Updating syn v1.0.99 -> v1.0.105
    Updating tempfile v3.2.0 -> v3.3.0
    Updating termtree v0.2.3 -> v0.4.0
    Updating textwrap v0.15.0 -> v0.16.0
    Updating thiserror v1.0.30 -> v1.0.37
    Updating thiserror-impl v1.0.30 -> v1.0.37
    Updating thread_local v1.1.3 -> v1.1.4
    Updating time v0.1.43 -> v0.3.17
      Adding time-core v0.1.0
      Adding time-macros v0.2.6
    Updating typenum v1.14.0 -> v1.16.0
    Updating unicode-ident v1.0.3 -> v1.0.5
    Removing unindent v0.1.7
    Updating uuid v1.0.0 -> v1.2.2
    Updating va_list v0.1.3 -> v0.1.4
    Updating version_check v0.9.3 -> v0.9.4
    Updating wasi v0.10.2+wasi-snapshot-preview1 -> v0.11.0+wasi-snapshot-preview1
```

NOTE: there's still a bug, because we set each new replicaset's weight
to 1 (even ones which don't satisfy replication factor)
before the bucket distribution is bootstrapped. But there must be at
least one replicaset with non zero weight in order for vshard.*.cfg to
work.

A potential solution would be to only configure vshard once a replicaset
is filled up.