Investigate replacing webpack with esbuild

While #328406 and !65170 (closed) look at using esbuild within our existing webpack tooling, this issue is about investigating what it would take/look like to replace webpack with esbuild in our build process.

The goal here is only to see what the theoretical fastest build process we can get from esbuild. In other words, this is strictly a proof-of-concept; it's unlikely this can currently be taken forward into a working build process.

The investigation would go something like this, with each step requiring the previous step to be "successful":

1. Build one of our entry points with esbuild to see if it's possible in principle (it is).
2. Build all of our entry points with esbuild to get a comparison of build time (a 2x2 matrix of dev/prod vs esbuild/webpack). If it's "significantly faster" (to be defined), proceed.
3. Build all entry points correctly, so that they work in a static file configuration for the GDK. This almost certainly isn't possible yet, due to limitations of esbuild (e.g., bundle splitting only working for esm output format)

The first and most simple step is done roughly in !73193 (closed). The second step will be more work, and the third step might be the most work, if it's even possible at all.