Autodump pickledb

[StaticallyTypedAnxiety]

Please check one of the following, relating to the CHANGELOG, which should be updated if relevant

• my changes to the code affect some exposed aspect of the developer experience, and I have added an item to relevant 'Added, Fixed, Changed, Removed, Deprecated, or Security' heading under the 'Unreleased' heading of the CHANGELOG, with the format - summary of change [PR#1234](https://github.com/holochain/holochain-rust/pull/1234)
• [x ] my changes to the code do not affect any exposed aspect of the developer experience

[StaticallyTypedAnxiety]

Switching to autodump policy dumps to the db everytime there is a change.

test bench_file_eav_many_to_one ... bench: 276,004,040 ns/iter (+/- 127,700,185)
test bench_file_eav_one_to_many ... bench: 286,338,120 ns/iter (+/- 142,272,851)
test bench_memory_eav_many_to_one ... bench: 1,607,075 ns/iter (+/- 2,202,613)
test bench_memory_eav_one_to_many ... bench: 1,355,982 ns/iter (+/- 406,924)
test bench_pickle_eav_many_to_one ... bench: 66,729,510 ns/iter (+/- 31,334,739)
test bench_pickle_eav_one_to_many ... bench: 64,792,690 ns/iter (+/- 31,317,922)

when running periodic policy it was a 92% increase, running this gives a 75% increase. Lose some performance but it is still good.

[lucksus]

I was looking into this already merged PR because I just encountered a new performance issue. In this latest FlameGraph that I just created after fixing the 100% CPU issue, PicklDB::dump is the biggest bar:

I guess your benchmark does not include running a node, and I also guess that in the context of a running instance the performance hit looks different..?

I would like to have this be revisited. I don't think we need nor should dump on every change.