Improve API for configuring the schema for a BigQuery Table

[quartzmo]

The current JSON-as-hash mechanism for setting a Table schema is needlessly verbose and exposes the all-caps convention of the underlying Rest API. Here is the example from the docs.

require "gcloud"

gcloud = Gcloud.new
bigquery = gcloud.bigquery
dataset = bigquery.dataset "my_dataset"

schema = {
  "fields" => [
    {
      "name" => "first_name",
      "type" => "STRING",
      "mode" => "REQUIRED"
    },
    {
      "name" => "cities_lived",
      "type" => "RECORD",
      "mode" => "REPEATED",
      "fields" => [
        {
          "name" => "place",
          "type" => "STRING",
          "mode" => "REQUIRED"
        },
        {
          "name" => "number_of_years",
          "type" => "INTEGER",
          "mode" => "REQUIRED"
        }
      ]
    }
  ]
}
table = dataset.create_table "my_table",
                             name: "My Table",
                             schema: schema

I believe it would improve the API to use something more structured and native to Ruby, such as a builder inside a block:

table = dataset.create_table "my_table", name: "My Table" do |schema|
  schema.string :first_name, mode: :required
  schema.record :cities_lived, mode: :repeated do |repeated_schema|
    repeated_schema.string :place, mode: :required
    repeated_schema.integer :number_of_years, mode: :required
  end
end

Both Dataset#create_table and Table#schema= would need to be updated for the new approach.

[jgeewax]

Likely related to googleapis/google-cloud-python#1058 (Python)