S3 methods troubles

[dajmcdon]

We've started to run into some of the S3 methods issues that @ryantibs described toward the bottom of #7. So far

• head() and tail() drop the class #96 has this behaviour specifically. It's not immediately clear that inheriting tsibble would fix the problem either, though it might.
• Looks like Make print.epi_df header mirror style of print.tsibble and print.tbl_df #113 has this as well, though less a bug and more an improvement.

I'm not yet suggesting that we revisit the decision in #7 (and changing now would be a big overhaul). For the moment, I'd like to use this issue to document any similar bugs. So if we find unexpected behaviour that occurs for standard data.frame processing, please create an issue and mention this one.