perf: Improve performance especially of validate and clone

[Uzlopak]

This PR improves the performance of especially validate and clone.

before:

MONGOOSE_DEV=1 node validate
invalid x 18,325 ops/sec ±1.40% (89 runs sampled)
valid x 46,240 ops/sec ±1.30% (91 runs sampled)

MONGOOSE_DEV=1 node clone
Simple x 124,034 ops/sec ±1.41% (91 runs sampled)
AllSchema x 17,409 ops/sec ±0.52% (94 runs sampled)

MONGOOSE_DEV=1 node create
CheckItem x 113,922 ops/sec ±0.34% (94 runs sampled)
Board x 112 ops/sec ±1.26% (74 runs sampled)

node populate
1436 completed queries on mongoose version 6.1.8

after:

MONGOOSE_DEV=1 node validate
invalid x 24,576 ops/sec ±0.89% (88 runs sampled)
valid x 119,331 ops/sec ±0.12% (95 runs sampled)

MONGOOSE_DEV=1 node clone
Simple x 148,691 ops/sec ±1.36% (87 runs sampled)
AllSchema x 22,479 ops/sec ±0.91% (95 runs sampled)

MONGOOSE_DEV=1 node create
CheckItem x 112,564 ops/sec ±0.26% (93 runs sampled)
Board x 122 ops/sec ±0.45% (77 runs sampled)

node populate
1528 completed queries on mongoose version 6.1.8
1528 completed queries on mongoose version 6.1.8

[Uzlopak]

I think the biggest performance bottleneck is the ES6 Proxy in DocumentArray and MongooseArray. Playing currently aroung with using WeakSets to see if the isMongooseArray and isMongooseDocumentArray can be boosted or not.

Also trying to figure out if we can reduce the roundtrips through the proxy layer by using forEach etc. when iterating through the elements of those special arrays.

[Uzlopak]

@vkarpov15
Should i split it into more granular PRs?

[vkarpov15]

Couple of comments, but general idea looks good

[vkarpov15]

This is a little risky because Mongoose clone() does a deep clone, so this may cause issues if there's a validator with nested objects

[vkarpov15]

LGTM, I'll just make the clone() change in master.