Make brotli optional

[siddharthkp]

Motivation and Context

From the looks of it, iltorb's native bindings are large + don't play well on all platforms - and fails installs on windows and some CI environments.

1. For folks not using brotli compression at all, this is unnecessary overhead.
2. Native brotli compression support landed in node 10.16.0 (LTS), we don't need to include an external dependency for that
3. For folks using brotli and node < 10.16.0, this PR is a breaking change and will ask them to install bundlesize-plugin-brotli to continue

Changes introduced in this PR:

1. Remove brotli-size from dependencies

2. On Node >= 10.6.0, use zlib.brotliCompressSync

3. On Node < 10.6.0, ask to install bundlesize-plugin-brotli

4. Introduces bundlesize-plugin-brotli

Fixes #202, #283 and #213

Prior work: #220 #299

[siddharthkp]

Open questions

• Should we merge this now or wait for 1.0.0 because of the breaking change?

[Haroenv]

I guess it depends how many things we want to change still in a v1, I think it's a breaking change as well

[siddharthkp]

I guess it depends how many things we want to change still in a v1, I think it's a breaking change as well

True, might make sense to keep them all in a branch and release together.

[anikethsaha]

bundlesize-plugin-brotli

@siddharthkp You are making bundlesize plugable? if so, its just AWESOME

Plugins are the kind of best suites for Scalability.

[siddharthkp]

Merged into beta channel

[thomaswelton]

It also looks like brotli size has been updated to drop iltorb
erwinmombay/brotli-size@8ca8a27

Is there a roadmap for release of your beta?

[siddharthkp]

Hi!

You might want to use this version of bundlesize that solves this issue: https://github.com/siddharthkp/bundlesize2#migration-from-bundlesize0180

it's not feature complete, but you might not need the missing features at all.