Support for chunk decoding on web workers

[kylebarron]

I'm reading up on zarrita.js and comparing it to https://geotiffjs.github.io/.

One of the nice user-facing amenities of geotiff.js is its Pool abstraction:

Using decoder pools to improve parsing performance

Decoding compressed images can be a time consuming process. To minimize this geotiff.js provides the Pool mechanism which uses WebWorkers to split the amount of work on multiple 'threads'.

const pool = new GeoTIFF.Pool();
const data = await image.readRasters({ pool });

It is possible to provide a pool size (i.e: number of workers), by default the number of available processors is used.

It doesn't seem that it's currently possible to do chunk decoding on web workers. A nice potential API would be to add some sort of pool to the get function.

This is also related to #296 about the comment:

with a fetch store, the requests are async but decompression is blocking

@aganders3 's solution was to implement their own web worker, but providing some sort of web worker pool API would mean that users could use main thread requests and not have to bundle their own workers.

[thewtex]

I made available getWorker and setWorker functions, providing web worker pool implementations for get and set, to complement zarrita.js in a package called fizarrita. Here is a benchmark app.

[kylebarron]

@manzt do you think it's in scope to have some sort of worker abstraction in zarrita.js itself?

[manzt]

I think this is interesting. I need to understand the primitives better. v0.7 is shaping up to be a bigger release and then I think this is something we can tackle in a followup.