perf: use ArrayVec for Stack

[mooori]

Changes the type of field Stack.data from Vec<U256> to ArrayVec<U256, 1024>.
This optimization was proposed in #448 and is based on the limit of 1024
items for the EVM stack.

Change of a public interface

Stack::data() returns &[U256] instead of &Vec<U256>.

NEAR gas usage

There a few tests in aurora-engine which fail due to gas usage lower than
expected:

---- tests::repro::repro_8ru7VEA stdout ----
thread 'tests::repro::repro_8ru7VEA' panicked at 'assertion failed: `(left == right)`
  left: `242`,
 right: `240`', engine-tests/src/tests/repro.rs:146:5
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- tests::repro::repro_D98vwmi stdout ----
thread 'tests::repro::repro_D98vwmi' panicked at 'assertion failed: `(left == right)`
  left: `199`,
 right: `198`', engine-tests/src/tests/repro.rs:146:5

---- tests::repro::repro_5bEgfRQ stdout ----
thread 'tests::repro::repro_5bEgfRQ' panicked at 'assertion failed: `(left == right)`
  left: `706`,
 right: `705`', engine-tests/src/tests/repro.rs:146:5

---- tests::repro::repro_FRcorNv stdout ----
thread 'tests::repro::repro_FRcorNv' panicked at 'assertion failed: `(left == right)`
  left: `197`,
 right: `196`', engine-tests/src/tests/repro.rs:146:5

---- tests::repro::repro_GdASJ3KESs stdout ----
thread 'tests::repro::repro_GdASJ3KESs' panicked at 'assertion failed: `(left == right)`
  left: `133`,
 right: `132`', engine-tests/src/tests/repro.rs:146:5

---- tests::uniswap::test_uniswap_input_multihop stdout ----
thread 'tests::uniswap::test_uniswap_input_multihop' panicked at 'assertion failed: `(left == right)`
  left: `123`,
 right: `122`', engine-tests/src/tests/uniswap.rs:41:5

Issue with jsontests

This change causes a stack overflow in evm-tests/jsontests. Identifying
the cause of a stack overflow in Rust can be tricky. Therefore, before
trying to get to the bottom of this, it might make sense to evaluate if this
change could/would be merged at all?

Note: It seems like the latest jsontests commit suitable for Aurora's EVM fork
is c66b562.

[birchmd]

Thanks for looking into this @mooori . In my opinion, it is probably not worth trying to figure out what goes wrong in the jsontests for a less than 1% gas savings. However, I do think it would be worth getting our fork up to date with the latest changes on the upstream sputnikvm project so that we can use the latest jsontests as well.

[mooori]

I will look into getting the fork up to date with upstream.