no more code duplication bw liMessage and rawMessage + several bug fixes

[timotheecour]

address leftover TODO from #14317

• no more code duplication bw liMessage and rawMessage (duplication since e254741 in 2009); these functions have accumulated odd unintentional (IMO) differences over the years and make patches harder to write and read; this merge should fix several bugs (or at least fragile bug-prone logic), eg
  • hintProcessing was treated differently for these (so rawMessage(graph.config, hintProcessing, s.name.s) was not rendering as dots); this didnt' seem intentional
  • hintMsgOrigin wasn't honored
  • warningAsError was treated suspiciously in rawMessage, and might've simply ignored warningAsError (eg with warnCannotOpenFile)
• fix what seems like a bug in msgWrite that would unconditionally reset lastMsgWasDot even if errorOutputs was empty (eg gagging during compiles())
• --hint:msgorigin should now work with all compiler generated messages
• internalAssert now formats instantiationInfo correctly
• removed assertNotNil (used once) + other code simplifications, resulting in negative diff stats

after PR

• I encountered other bugs while fixing this but best to keep this one as mostly refactoring, next one will fix bugs I've encountered (at least they'll only need to be fixed in 1 place)