Changes in protobuf.js 6.0

• Instead of relying on bytebuffer.js, there is now a use case specific and reusable wire format reader / writer interface that abstract away any differences between node and modern to ancient browsers.
• Overall minified size is now ~50kb instead of ~80kb with bytebuffer. ~15kb gziped.
• Basic long support has been integrated into the library to fall back to (possibly unsafe) JavaScript numbers if no long library is present.
• Builder functionality has been integrated into reflection. Instead of calling Builder#build, there is now Namespace#lookup which returns the reflected type for a path.
• Loading .proto or .json files is now all asynchronous and loads all files in parallel.
• Creating message instances from reflected types is done through calling MyType#create instead of building classes and calling new. Additionally, plain objects can now be encoded just like runtime messages and you can now inherit from the internal prototype to create your own classes with custom functionality.
• Runtime messages don't have encode and decode methods anymore by default. These are now static methods on the reflected type respectively custom constructor and can be used with both actual runtime messages and plain objects.
• There are no more null values marking unset fields. Instead, default values are on the prototype now and existence on the wire can be checked with Object#hasOwnProperty.
• Reflection has built-in code generation which makes encoding and decoding of messages blazingly fast.
• Common Google types within any.proto, duration.proto, timestamp.proto, empty.proto and struct.proto are now integrated.
• The build-process has been changed to gulp/browserify for seamless integration into your own project.
• Documentation has received a major overhaul.
• TypeScript definitions are now automatically generated and the API is built in a way that makes it easier to work with types dialects (i.e. working with typed reflection objects instead of just magically built messages).
• The pbjs CLI now sets up itself only when required. Its dependencies are not pulled on npm install.
• It's now easier to generate static code. There is actually a static target for pbjs now as a starting point.