Fix loop validation input context is mutable

[Maks3w]

Given two input fields the context used in each one will be different. https://github.com/zendframework/zend-inputfilter/pull/24/files#diff-2cd0c1ba642dc3e6de9ca4e2af0f5e71R290

This means inputs need to be order sensitive

$inputs = ['A', 'B'];
$data = [ /* empty */ ];

$inputA->isValid($context = ['A' => null]);
$inputB->isValid($context = ['A' => null, 'B' => null]);

// Inverse order will change $context too.
$inputs = ['B', 'A'];
$data = [ /* empty */ ];

$inputB->isValid($context = ['B' => null]);
$inputA->isValid($context = ['B' => null, 'A' => null]);

This PR provides fix and enhance this feature by building a default input context using the default input values when no custom context is provided.

Fix #16

Supersede and close #30

[Maks3w]

Note tests are not failling just the coverage due the different weights given to each line after removal.

[svycka]

looks good
also have to note that this adds something new

$inputs = ['A', 'B'];
$data = ['C' => 'something'];

In context now we have:

['A' => null, 'B' => null, 'C' => 'something']

Before 'C' => 'something' would not be added to context.
But I think this is not a bug but good feature. Maybe add test to keep it :)

[Maks3w]

Yep I opt for use default input raw value for build the context (usually null, sometimes mixed). Test it's placed for this enhancement.

[weierophinney]

As I noted on #25, I'm a bit worried that changing test assumptions is indicative of a BC break; also, like #25, the logic looks reasonable. I'll do a more thorough review early next week.

[Maks3w]

May this should use the filtered version instead the raw value

[weierophinney]

Considering that the original from two lines above was using the raw values, we should keep it that way for now. We can address again with the next major version.

[Maks3w]

Note before the context was always null for optional values. getRawValues already change the behavior if user call setValue before

[weierophinney]

@Maks3w Can you rebase this, please? When I merged, there was a conflict, and my attempts to resolve the conflict have left me with failing tests.

[Maks3w]

Since #25 optional fields is not call what was expected. I've to rewrite the test

[Maks3w]

Done. Now includes changes from #30

[weierophinney]

This looks good per our discussion last week on IRC.

For those reading:

• If the value was not present but has a fallback, we set it to the default value and consider it valid.
• If the value was not present, but required, we mark it as invalid.