Improve implementation of Freeze and Release

[ktbarrett]

Currently, these actions are implemented using the same mechanism as regular value sets. They first sample the current value and then use then as the value in their "set". This overwrites other explicit writes with the implicit write the implementation does. This is surprising behavior and should be avoided.

If you update a value and then immediately do a Freeze, it will freeze the current simulator value, not the value just scheduled.

# dut.signal.value == 10
dut.signal.value = 12
dut.signal.value = Freeze()
await Timer(...)
assert dut.signal.value == 12  # will always fail

Or if you do a Force followed immediately followed by a Release you get the same issue

# dut.signal.value == 10
dut.signal.value = Force(12)
dut.signal.value = Release()
await Timer(...)
assert dut.signal.value == 12  # will always fail