Implement SAX::Parser for cobertura parser

Problem

Parsing big (> 300MB) XML cobertura files could blow up memory.

We already discussed about this potential ~performance improvement last year, and today might be a good time to reevaluate our strategy.

Proposal

From my current understanding, our current implementation is parsing the XML file at once.

By leveraging a SAX::PARSER we could iterate through the XML instead of loading everything as once leading to ~performance improvement.

We could also implement this solution for our Junit parser in a next iteration.

Below some basic code that we could take inspiration:

require 'nokogiri'
require 'memory_profiler'

class MyDocument < Nokogiri::XML::SAX::Document
  def end_document
    puts "the document has ended"
  end

  def start_element name, attributes = []
#    puts "#{name} started"
  end
end

parser = Nokogiri::XML::SAX::Parser.new(MyDocument.new)

report = MemoryProfiler.report do
  parser.parse(File.open('cobertura-coverage.xml'))
end

report.pretty_print

cc @drew @iamricecake @jheimbuck_gl @shampton