Use be_empty expectations in more specs

[HertzDevil]

This patch replaces most usages of empty?.should be_true with should be_empty, and empty?.should be_false with should_not be_empty, because be_empty's failure message is more descriptive:

require "spec"

it { "".should_not be_empty } # preferred
it { "".empty?.should be_false }
it { "".empty?.should_not be_true }
it { "a".should be_empty } # preferred
it { "a".empty?.should be_true }
it { "a".empty?.should_not be_false }

FFFFFF

Failures:

  1) assert
     Failure/Error: it { "".should_not be_empty }

       Expected: "" not to be empty

     # test.cr:3

  2) assert
     Failure/Error: it { "".empty?.should be_false }

       Expected: false
            got: true

     # test.cr:4

  3) assert
     Failure/Error: it { "".empty?.should_not be_true }

       Expected: actual_value != true
            got: true

     # test.cr:5

  4) assert
     Failure/Error: it { "a".should be_empty }

       Expected: "a" to be empty

     # test.cr:6

  5) assert
     Failure/Error: it { "a".empty?.should be_true }

       Expected: true
            got: false

     # test.cr:7

  6) assert
     Failure/Error: it { "a".empty?.should_not be_false }

       Expected: actual_value != false
            got: false

     # test.cr:8

The only exceptions are specs testing the #empty? method itself. Specs like these are preserved:

crystal/spec/std/uri/params_spec.cr

Lines 299 to 305 in 458595f

	describe "#empty?" do
	it "test empty?" do
	Params.parse("foo=bar&foo=baz&baz=qux").empty?.should be_false
	Params.parse("").empty?.should be_true
	Params.new.empty?.should be_true
	end
	end

[straight-shoota]

I would even consider be_empty for testing #empty? explicitly.