Fixed Replacing a value with 0 with dataframe.replace throws an error

[junduck]

Fixes #621

The original implementation wrongly tests falsy on oldValue and newValue, which results in throwing when replacing with falsy values like 0 or empty string.

This fix test for undefined instead, then validates oldValue/newValue types.

This fix is trivial so no test coverage.

• Previous PR was closed due to editor removal of whitespace that spamming diff log