[optimisation] quick win - byte size optimised address mask

## Abstract and Motivation

Solidity uses bitmask (0xffffffffffffffffffffffffffffffffffffffff) to sanity clean address value.

It compiles that into this code:

```
PUSH20 0xffffffffffffffffffffffffffffffffffffffff
```

This requires 21 bytes and 3 gas for each situation address is used. (70+ times in some example ERC20 bytecode).

I propose much shorter approach for the case the optimiser is set to optimise bytecode length.

## Solution

Compile to this instead (in the case the optimiser is set to optimise bytecode length):
```
PUSH0
NOT
PUSH1 12
SHR
```

This results into 5 bytes code and 11 gas.

In some common ERC20 bytecode it has impact on being 1100+ bytes shorter. Wrt to functions - there can be just +40 gas needed in the case the address is sanity cleaned 5 times within 1 call.

## Backwards Compatibility

The optimised bytecode is shorter (-16 bytes), but has +8 gas. Functionally the same.

The issue names no file, test, or entry point; begin by locating the compiler code that emits the address mask and the optimizer's bytecode-length mode. Compare generated bytecode and gas for the proposed sequence against the current mask, including the ERC20 example; done means the shorter output is selected only for that mode while functional behavior remains unchanged.

Written by the indexing model from the issue text.