Fixes #307.

Notably this support relies on string to identifier conversion, which
works natively using Ltac2 in Coq 8.11+ and with a plugin
(https://github.com/ppedrot/coq-string-ident) in Coq 8.10. To use it,
you must replace intro_patterns.string_to_ident_hook with a real
implementation; see https://gitlab.mpi-sws.org/iris/string-ident for a
working implementation that works with Coq 8.11 (using Ltac2).

The syntax is %H (within a string intro pattern). This is technically
backwards-incompatible, because this was previously supported and parsed
as % and H separately. To restore the old behavior, separate with a
space, eg [% H].

- The error messages were wrong: the goal needs to be absorbing, not the hypothesis.
- The wrong failure number was used in `iAssumption`, which caused the error not
  to be propagated properly.

Gregory Malecha authored 5 years ago and Robbert Krebbers committed 5 years ago

- remove "odd" comment
- move atomic triples to bi_scope

Paolo G. Giarrusso authored 5 years ago and Robbert Krebbers committed 5 years ago

When proving `foo` through a fixpoint, Coq's guardedness checker needs to see to
which arguments `foo` is applied. Opaque lemmas applied to `foo` itself prevent
that, so make them transparent.
* Make `IntoEmpValid` lemmas transparent.
* Expose application of `IntoEmpValid` instance to its argument.
* Add comment to `tac_pose_proof`

This MR brings back the type of `tac_pose_proof` to the one it had before !329.
Hence, this seems worth a comment.

This test case arose in Iron.

Also, rewrite `iIntoEmpValid`. Now, instead of using Ltac to traverse
the type of the term and generate goals for the premises, we repeatedly
apply a series of lemmas. This has the advantage that it works up to
convertability, and we no longer need the `eval ...` hacks.

Fix issue pointed out by @tchajed in !330 (comment 41322)