Refactor: handle view encoding (Uint8Array) natively

[vweevers]

Replaces the internal asBuffer option with an encoding option that can be one of buffer, utf8 or view, and converts data accordingly. No longer need to transcode view to buffer.

Has no performance effect on get() with utf8, buffer or view encoding:

Click to expand

Same for iterator() (main is slightly faster at utf8 in this graph, but on average there's no difference):

Click to expand

[vweevers]

I also tried napi_create_typedarray() which I expected to be faster for skipping a copy but it's actually a lot slower... Same goes for Buffer when using napi_create_external_buffer() instead of napi_create_buffer_copy(). I didn't dive into why it's slower, so here's the code for future reference:

void Convert2 (napi_env env, char* data, const size_t size, const Encoding encoding, napi_value& result) {
  if (data == NULL) {
    napi_get_undefined(env, &result);
  } else if (encoding == Encoding::buffer) {
    napi_create_external_buffer(env, size, data, NULL, NULL, &result);

    // Delete data once ArrayBuffer is garbage collected
    napi_value arrayBuffer;
    napi_get_named_property(env, result, "buffer", &arrayBuffer);
    napi_wrap(env, arrayBuffer, data, FinalizeArrayBuffer, NULL, NULL);
  } else {
    napi_value arrayBuffer;
    napi_create_external_arraybuffer(env, data, size, FinalizeArrayBuffer, NULL, &arrayBuffer);
    napi_create_typedarray(env, napi_typedarray_type::napi_uint8_array, size, arrayBuffer, 0, &result);
  }
}

void FinalizeArrayBuffer (napi_env env, void* data, void* hint) {
  if (data) {
    delete (char*) data;
  }
}