Merge with Upstream and Simplify Decoding

[willemneal]

• Decreases binaries by ~50% by defining fewer classes
• Handle Type Aliases
• Fixes stub allocation
• Loader much faster at accessing typed arrays
• Fixes typing generation

[willemneal]

It fails the tests because the new loader changed APIs.

[vgrichina]

@willemneal LGTM, but let's create demo fiddle on studio (overriding compiler in setup.js) to make sure this works

[willemneal]

Okay I'll make one now. Could we set this up to use the render bot?

[willemneal]

https://near-studio-staging-pr-154.onrender.com/?f=zftjgvt4y&asc=3edc33aa363b49147ab783111f9940879eae0a95

This uses the new URL asc parameter from https://github.com/nearprotocol/NEARStudio/pull/154

But here is this commit on the live site: https://studio.nearprotocol.com/?f=p8h1iu1hq

[bowenwang1996]

Do you mind pulling the latest upstream again? The resolver fix would be quite useful.

[janedegtiareva]

Do you mind merging master so that we can easily see results of vm tests?

[willemneal]

Sure just updating merging with upstream and there were some issues introduced by the second to last PR so I'm going to merge the one before it.

[bowenwang1996]

What was the issue?

[willemneal]

Not exactly sure. The errors are internal to bignum and it no longer likes the way I handled u128 as a special case. I tried rolling back to just the PR with type resolution and I get the same thing. I'm now trying the PR before to see if it is what introduces this error.

[willemneal]

So I narrowed down the error from the type resolution PR to bignum. I just created a PR there: https://github.com/MaxGraey/bignum.wasm/pull/27 to fix it. For now I just merged with the PR before it and we can return after bignum is fixed.

[willemneal]

@vgrichina I also changed the API so that obj.encode() returns a Uint8Array and so does encode<Obj>(obj).