When the interest cache is configured as part of initing the `LogTracer` before
the global dispatcher is set, the main thread fails to log. The problem
is that the interests cache registers a dummy callsite. With
tokio-rs#2083 the callsite registration
changed to use `dispatcher::get_default`. This call internally sets the
thread local `default` dispatcher to `Dispatch::none` when there isn't
yet any dispatcher set. This `default` dispatcher isn't overwritten when
setting the global dispatcher. The solution is to not store the
`Dispatch::none` as `default` dispatcher.

this brings performance of most of the `subscriber.rs` benchmarks closer
to where they were on `v0.1.x` prior to this change (commit
cf98b73):

```
     Running unittests (target/release/deps/subscriber-b3c6395267c4f6e6)
Gnuplot not found, disabling plotting
span_no_fields          time:   [18.814 ns 18.900 ns 18.967 ns]
                        change: [-17.562% -17.045% -16.552%] (p = 0.00 < 0.05)
                        Performance has improved.

enter_span              time:   [2.8471 ns 2.8620 ns 2.8799 ns]
                        change: [+7.5923% +8.2941% +8.9390%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 20 outliers among 100 measurements (20.00%)
  2 (2.00%) low severe
  2 (2.00%) high mild
  16 (16.00%) high severe

span_repeatedly         time:   [3.6959 us 3.6988 us 3.7017 us]
                        change: [-3.0008% -2.9276% -2.8528%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 100 measurements (1.00%)
  1 (1.00%) high mild

span_with_fields        time:   [56.854 ns 56.911 ns 56.975 ns]
                        change: [-3.5756% -3.4955% -3.3997%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 20 outliers among 100 measurements (20.00%)
  1 (1.00%) low mild
  3 (3.00%) high mild
  16 (16.00%) high severe

span_with_fields_record time:   [317.15 ns 318.02 ns 318.94 ns]
                        change: [+3.1700% +3.3617% +3.5555%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 9 outliers among 100 measurements (9.00%)
  4 (4.00%) high mild
  5 (5.00%) high severe

dispatch/no_dispatch_get_ref
                        time:   [17.823 ns 17.966 ns 18.134 ns]
                        change: [-10.826% -10.144% -9.4796%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  3 (3.00%) high mild
dispatch/no_dispatch_get_clone
                        time:   [24.984 ns 25.013 ns 25.041 ns]
                        change: [-6.7511% -6.6349% -6.5020%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 8 outliers among 100 measurements (8.00%)
  1 (1.00%) low severe
  2 (2.00%) low mild
  1 (1.00%) high mild
  4 (4.00%) high severe
dispatch/get_ref        time:   [3.8508 ns 3.8532 ns 3.8555 ns]
                        change: [-26.641% -26.518% -26.404%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  1 (1.00%) low mild
  2 (2.00%) high mild
  3 (3.00%) high severe
dispatch/get_clone      time:   [10.737 ns 10.750 ns 10.765 ns]
                        change: [-11.632% -11.429% -11.219%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 17 outliers among 100 measurements (17.00%)
  3 (3.00%) low mild
  2 (2.00%) high mild
  12 (12.00%) high severe

Gnuplot not found, disabling plotting
```

performance of `get_default` where there is no current subscriber is
still significantly regressed vs `v0.1.x`, due to `Arc` allocations for
`Dispatch::none` occurring every time. i think this can be solved as well.

Signed-off-by: Eliza Weisman <eliza@buoyant.io>

It also rewrites the logic on how the `current` dispatcher is determined.