Memory allocated in vectorized operations

This is an extension of issue #121. While individual accesses and sets have been fixed, vectorized operations are still allocating significantly more than they should. Here's a MWE:

```julia
using ACSets

SchTest = BasicSchema([:V,:E], [(:v0,:E,:V)])

@acset_type Test(SchTest, index=[:v0])

s = Test()

N = 10
add_parts!(s, :V, N)
idx = add_parts!(s, :E, N)

entry = ones(Int64, N);

function add_vector(s, idx, entry)
s[idx, :v0] = entry
end

function add_loop(s, idx, entry)
for (i, id) in enumerate(idx)
s[id, :v0] = entry[i]
end
end

begin
add_vector(s, idx, entry)
add_loop(s, idx, entry)
end

@info "Using add_vector"
@time add_vector(s, idx, entry)

@info "Using add_loop"
@time add_loop(s, idx, entry)
```

As well as the results: