Add gem startup benchmarking

[marcotc]

Similar to #1043, this PR adds memory and timing benchmarking to the gem loading process: require 'ddtrace'.

Here's the sample output:

ddtrace gem memory footprint: 6199.06 KiB
ddtrace gem load time: 0.155863s

We run the gem loading process a few times in a new process (to be able to require again like it is the first time) and collect the average to create these results.

Plus, we also have the detailed memory report when running locally:

Total allocated: 6477394 bytes (47492 objects)
Total retained:  895123 bytes (7032 objects)

allocated memory by gem
-----------------------------------
   6455955  ./lib
     21439  other

allocated memory by file
-----------------------------------
    685735  ./lib/ddtrace.rb
    276785  ./lib/ddtrace/tracer.rb
    200023  ./lib/ddtrace/configuration/options.rb
    174986  ./lib/ddtrace/span.rb
    154700  ./lib/ddtrace/transport/http/adapters/unix_socket.rb
    139054  ./lib/ddtrace/configuration/option_definition.rb
    125707  ./lib/ddtrace/runtime/socket.rb
    125132  ./lib/ddtrace/transport/http.rb
     91641  ./lib/ddtrace/writer.rb

[brettlangdon]

cool!