Improve blob reader performance

[richarddd]

Description of changes:
Improve performance of collecting blob body by usingarrayBuffer()method.

If one or more of the packages in the /packages directory has been modified, be sure yarn changeset add has been run and its output has
been committed and included in this pull request. See CONTRIBUTING.md.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[kuhe]

could you update the unit tests for this?

what kind of performance diff does this bring?

[richarddd]

could you update the unit tests for this?

Sure

what kind of performance diff does this bring?

It will depend on the payload size. But here we're using native (c++) apis to asynchronously populate an array buffer from the read stream in one go.

In the current implementation we do all these steps:

1. we first have to base 64 encode the stream to a string (which requires a lot of memory).
2. Then run a regex on the string to check its a valid base64 string
3. Then do all this on each 4 bytes

   smithy-typescript/packages/util-base64/src/fromBase64.browser.ts

   Line 10 in 9ff2027

   export const fromBase64 = (input: string): Uint8Array => {

So at least 2-5x is very pessimistic (for browsers).
For node is not as dramatic, but we do run a regex on the entire string to validate its base64.

Then we also have this code that does 95% the same thing:

smithy-typescript/packages/chunked-blob-reader-native/src/index.ts

Line 5 in 9ff2027

export function blobReader(

Its out of scope for this PR but in fact, all base64 decoding that is being used to ultimately return a typed array/byte array can be removed.

[kuhe]

https://caniuse.com/mdn-api_blob_arraybuffer checks out, and for Node.js if we have Blob (v18) global then it has .arrayBuffer() on it.