Use NonNull in ReleasePool and add bench_dict

[kngwyu]

1st step to #271

[konstin]

Test should use new(...).unwrap() instead of new_unchecked(...), otherwise 👍

[kngwyu]

@konstin
Travis fails with

running 5 tests
test binary_arithmetic ... ok
test inplace_operations ... FAILED
test rich_comparisons ... ok
test rich_comparisons_python_3_type_error ... ok
test unary_arithmetic ... ok
failures:
---- inplace_operations stdout ----
thread 'inplace_operations' panicked at 'attempt to subtract with overflow', tests/test_arithmetics.rs:76:9
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
stack backtrace:
   0: std::sys::unix::backtrace::tracing::imp::unwind_backtrace
             at libstd/sys/unix/backtrace/tracing/gcc_s.rs:49
   1: std::sys_common::backtrace::_print
             at libstd/sys_common/backtrace.rs:71
   2: std::panicking::default_hook::{{closure}}
             at libstd/sys_common/backtrace.rs:59
             at libstd/panicking.rs:211
   3: std::panicking::default_hook
             at libstd/panicking.rs:221
   4: std::panicking::rust_panic_with_hook
             at libstd/panicking.rs:476
   5: std::panicking::continue_panic_fmt
             at libstd/panicking.rs:390
   6: rust_begin_unwind
             at libstd/panicking.rs:325
   7: core::panicking::panic_fmt
             at libcore/panicking.rs:77
   8: core::panicking::panic
             at libcore/panicking.rs:52
   9: <test_arithmetics::InPlaceOperations as pyo3::class::number::PyNumberProtocol<'p>>::__isub__
             at tests/test_arithmetics.rs:76
  10: <T as pyo3::class::number::PyNumberISubProtocolImpl>::nb_inplace_subtract::wrap
             at ./src/class/macros.rs:153
  11: PyNumber_InPlaceSubtract
             at Objects/abstract.c:1054
             at Objects/abstract.c:1068
             at Objects/abstract.c:1087
  12: _PyEval_EvalFrameDefault
             at Python/ceval.c:1491
  13: _PyEval_EvalCodeWithName
             at Python/ceval.c:3971
  14: PyEval_EvalCodeEx
             at Python/ceval.c:4000
  15: PyEval_EvalCode
             at Python/ceval.c:522
  16: PyRun_StringFlags
             at Python/pythonrun.c:1035
             at Python/pythonrun.c:959
  17: pyo3::python::Python::run_code
             at src/python.rs:197
  18: pyo3::python::Python::run
             at src/python.rs:168
  19: test_arithmetics::inplace_operations::{{closure}}
             at tests/test_arithmetics.rs:118
  20: test_arithmetics::inplace_operations
             at tests/test_arithmetics.rs:122
  21: test_arithmetics::inplace_operations::{{closure}}
             at tests/test_arithmetics.rs:112
  22: core::ops::function::FnOnce::call_once
             at /rustc/6f93e93af6f823948cc13d2938957757c6486d88/src/libcore/ops/function.rs:238
  23: <F as alloc::boxed::FnBox<A>>::call_box
             at libtest/lib.rs:1471
             at /rustc/6f93e93af6f823948cc13d2938957757c6486d88/src/libcore/ops/function.rs:238
             at /rustc/6f93e93af6f823948cc13d2938957757c6486d88/src/liballoc/boxed.rs:673
  24: __rust_maybe_catch_panic
             at libpanic_unwind/lib.rs:102

But I can't reproduce this failure in my local machine.
Do you have any idea?

[konstin]

I broke the build because I didn't realize that (self.ptr as *mut u8) isn't the same as self.ptr when applying an offset. Should hopefully be fixed now.