Better v-table storage

[p-avital]

1.78 has introduced a regression that prevents stabby from implementing trait-object construction as efficiently as before.

This issue tracks progress in mitigating this problem into core::marker::Freeze is stabilized.

[xuchaoqian]

Benchmark feedback:

1. I constructed a use case as follows:

host ---> stabby plugin ---> host
The host asynchronously calls stabby plugin's API, which asynchronously callbacks the host.
vs.
host ---> host ---> host
The host asynchronously calls its own API, which asynchronously callbacks the host.

The goal is to compare stabby's async approach versus the native async approach in both the nightly and stable channels.

2. Here is the result:

3. Conclusion:

1) Stabby in the nightly channel is 20%+ faster than stabby in the stable channel.
2) Stabby's future implementation is 2-3 times slower than the native async implementation.

If you're interested, you can check the code here: stabby-plugin-demo