Customized aggregator for different edge attributes in to_simple transform & "max" aggregator

## 🚀 Feature

I suggest allowing users to choose different aggregators for different edge attributes. "max" aggregator keeps the largest value. The new function declaration would look like:
```
def to_simple(g,
return_counts='count',
writeback_mapping=False,
copy_ndata=True,
copy_edata=False,
etypes=None,
aggregator="arbitrary"):
"""
...
etypes: List[str], optional
Specifies the edge types to apply on. If None, merge parallel edges of all edge types.
aggregator: Union[Dict[str, str], str] optional
Indicates how different edge attributes are aggregated. If a string is provided, apply on all attributes. If a dict is provided, apply 'value' agrgegator to 'key' attribute.
"""
```

## Motivation

Some graphs may have parallel edges with different timestamps. In the case where the users need only the latest one, to_simple(aggregator="max") may be handy. Another common case is when the users need to merge parallel edges of only some specific types.