Split numbers embedded in non-snake-case words

[D-Walther]

Reasoning:

• Numbers in camelCase (e.g. camelCase123) in C are usually translated to defines by splitting them first (e.g. BYTE_SIZE_CAMEL_CASE_123). In this case, splitting numbers from letters makes sense.
• This split is not always sensible however. If a message is named MyMessage_mk2, the user has implicitly expressed that "mk" and "2" belong together, making "BYTE_SIZE_MY_MESSAGE_MK_2" an unintuitive result. An underscore in a word can therefore be interpreted as the user taking charge of the process; we should not interfere. Hence the "_" in word condition.
• This alone can result in inconsistencies in places where "UPPER_SNAKE_CASE" is used, however (e.g. in case of enum values and constants). If an enum value named "TI82_PLUS" results in "ti82_plus", "TI82" should not result in "ti_82". To remedy this, I've also excluded uppercase words.

In other words: if the user breaks convention, assume they have a reason for it. This includes:

• underscore in PascalCase or camelCase (e.g. MyMessage_v1, myMessage_v1 -> my_message_v1)
• numbers next to words in snake_case and UPPER_CASE (e.g. my_value1, MY_VALUE1 -> my_value_v1)