feat: support natspec annotations

[daejunpark]

fix #131

Support @custom:halmos natspec annotation to specify additional commandline options.

The natspec for a contract is inherited by every function in the contract. The natspec for a function can overwrite the inherited natspec.

For example, given:

/// @custom:halmos --opt1 --loop 1
contract C {

    /// @custom:halmos --opt2 --loop 2
    function checkA() { ... }
}

The options for the checkA() function are: --opt1 --opt2 --loop 2

Note that it inherits --opt1 but overwrites --loop 1.

--

The halmos natspec can be provided in multiline:

/// @custom:halmos --x
///                --y

or, using multiple tags:

/// @custom:halmos --x
/// @custom:halmos --y

[karmacoma-eth]

I was looking for something to avoid the need for a deepcopy here, I ended up with this:

import argparse
import collections

parser = argparse.ArgumentParser()
parser.add_argument('--foo')
parser.add_argument('--bar')
parser.add_argument('--loop')

def dictify(args):
    return {k: v for k, v in vars(args).items() if v is not None}

defaults = parser.parse_args(['--foo', 'beep', '--bar', 'boop'])
overrides = parser.parse_args(['--foo', 'ding', '--loop', '42'])
combined = collections.ChainMap(dictify(overrides), dictify(defaults))

print(combined['foo']) # 'ding' comes from overrides, takes precedence over defaults
print(combined['bar']) # 'boop' comes from defaults
print(combined['loop']) # '42' comes from overrides, not specified in defaults

ChainMap is cool but in the end it's not necessarily better because there is still some copying involved because of a poor interaction with Namespace. Might make sense to write our own NamespaceChain-like wrapper that avoids any copying at all at a later point

[daejunpark]

do you suggest using chainmap now, or doing something better later?

i wasn't super happy with using deepcopy here, but thought that the namespace object should be small enough and deepcopy wouldn't affect performance much. are there any other issues that i missed?

[karmacoma-eth]

no it's just a general dislike of deepcopy creep. Disregard for now

[karmacoma-eth]

this is great, can't wait to start using it

[daejunpark]

the external test failures are independent from this pr. it will be fixed by #134. will merge this pr after merging that.