Fix TaskArguments#deconstruct_keys with keys = nil

[nevans]

This fixes a bug in the initial implementation (#515):

#deconstruct_keys should handle keys == nil, or **rest will be broken.

For example, the normal behavior looks like this:

{a: 1,b: 2,c: 3}=>{a:, **rest}a# => 1rest# => {b: 2, c: 3}

But without handling keys == nil, we'll get this:

classExampledefinitialize(hash)=@hash=hashdefdeconstruct_keys(keys)=@hash.slice(*keys)endExample.new({a: 1,b: 2,c: 3})=>{a:, **rest}# !> "#{inspect}: key not found: :a" (NoMatchingPatternKeyError)

[rgarner]

Suggested change

	assert_equalta.deconstruct_keys(nil),{a: 1,b: 2,c: 3}
	assert_equalta.deconstruct_keys([:a,:b]),{a: 1,b: 2}
	assert_equal({a: 1,b: 2,c: 3},ta.deconstruct_keys(nil))
	assert_equal({a: 1,b: 2},ta.deconstruct_keys([:a,:b]))

Being an RSpec devotee, I spoke with an accent when I added this test and had the assert_equal(expected, actual) order backwards. It'd only affect the failure message but I'd hate for someone plugging a hole to copy my bad accent...

(I went to fix it but then noticed you'd opened this PR! Too quick for me 😉)

[nevans]

I'm also an rspec guy, so I get the order of these backwards pretty often myself. I certainly didn't notice they were in the wrong order here! 😉 Anyway, I pushed an update with them in the correct order. Thanks. 🙂