[soltest] Introduce explicit alignment

[erak]

As it turned out, we implimented the alignment of hex number literals in a confusing way (#6120 (comment)). This PR does the following two things:

1. Changes hex number literals (e.g. 0x23) such that they are right-aligned (which should be the natural behavior.)
2. Introduces two new keywords: left and right. These can be used to explicitly define the alignment of values:

contract C {
    uint256 public stateDecimal = 0x20;
}

contract D {
    bool public stateBool = true;
    uint256 public stateDecimal = 42;
    bytes32 public stateBytes = "\x42\x00\xef";

    function internalStateDecimal() public returns (uint256) {
        return (new C()).stateDecimal();
    }

    function update(bool _bool, uint256 _decimal, bytes32 _bytes) public returns (bool, uint256, bytes32) {
        stateBool = _bool;
        stateDecimal = _decimal;
        stateBytes = _bytes;
        return (stateBool, stateDecimal, stateBytes);
    }
}
// ----
// stateBool() -> true
// stateBool() -> right(true)
// stateDecimal() -> 42
// stateDecimal() -> right(42)
// stateBytes() -> left(0x4200ef)
// internalStateDecimal() -> 0x20
// update(bool,uint256,bytes32): false, -23, left(0x2300ef) -> false, -23, left(0x2300ef)

[codecov]

Codecov Report

Merging #6183 into develop will decrease coverage by 0.05%.
The diff coverage is 91.14%.

@@             Coverage Diff             @@
##           develop    #6183      +/-   ##
===========================================
- Coverage    87.97%   87.92%   -0.06%     
===========================================
  Files          377      377              
  Lines        36221    36191      -30     
  Branches      4286     4279       -7     
===========================================
- Hits         31867    31822      -45     
- Misses        2904     2913       +9     
- Partials      1450     1456       +6

Flag	Coverage Δ
#all	87.92% <91.14%> (-0.06%)	⬇️
#syntax	27.04% <5.29%> (-0.06%)	⬇️

[erak]

This is an important change that was needed to get the update in isoltest right if left() or right() was used to define the expected result.

[ekpyron]

One suggestion which can be addressed in varying intensity ;-). Otherwise this should be squashed (looking at the individual commits first was a bit misleading ;-)), but generally looks good to me!

[ekpyron]

You could probably inline alignRight (and alignLeft below). I also wonder whether skipping the whole DeclaredAlignment business and using a capture-less alignment lambda instead, supplemented by a hasAlignModifier flag for the alignment != Alignment::None check below might be nicer. Probably fine either way, though.
Also you could pass in a reference to ABIType and set its align member in the lambda instead of returning a pair - that might spare you reassigning ABIType::align over and over below. For that matter it could also just be a ref to bytes to which the padding is appended/prepended.

[erak]

Squashed and updated. @ekpyron Took in your suggestions.

[erak]

Argh, I put in quite some effort to make Codecov happy. But we're still testing tests here, so I think a delta of -0.03% should be fine ;)