Various fixes

• Removed the use of pyepics and converted everything to instead (only) use p4p
• Got rid of a lot of unnecessary run-iocsh wrapping and asserts related to require functionality and e3 infrastructure
• Added linting and fixed all issues except for the presence of magic numbers
• Various other readability and code quality improvements

Note: The original test had a bug where it compared decimated array elements to themselves (dec_array_y[-1] == dec_array_y[-1]) which would always pass. The fix compares to the original array elements, which is more meaningful but revealed that LTTB downsampling doesn't preserve middle elements as the test expected. The middle element assertion has been removed since LTTB is designed to preserve overall shape and endpoints, not specific index correspondences.