Some proposed backports from master to emacs-26

[Noam Postavsky]

I pushed some proposed backports to scrach/np/backport-26.2, listed here
by bug:

- #30931 abort() due to CHECK_ALLOCATED_AND_LIVE failure during GC
  Not a regression, but IMO nasty enough to be worth backporting.
  - Centralize Bug#30931 fix
  - Fix another case of freed markers in the undo-list (Bug#30931)
  - Fix CHECK_ALLOCATED_AND_LIVE abort during GC

- #30995 Slow window and frame popup on Emacs 26+ (e.g. with Helm)
  Regression.
  - Don't wait for visible frames to become visible

- bug#27656: 26.0.50; Fix "Make mail-extract-address-components return the user 
name more"
  Regression.
  - Revert "Make mail-extract-address-components return the user name more"

- bug#31146: 27.0.50; ielm via cl-prin1 infloops when printing circular object
  Bug in new feature.
  - Fix cl-print for circular sublists (Bug#31146)

- #30363 26.0.91; epa-decrypt-file not using epa-pinentry-mode
  Fix is trivial, and obviously safe.  Not a regression, but with
  removal of pinentry.el, epa-pinentry-mode becomes more important.
  - * lisp/epa.el (epa-decrypt-file): Apply epa-pinentry-mode (Bug#30363).

- part of Bug#28323 improper parsing of first argument
  Regression in Emacs 24.4
  - esh-opt.el: Fix improper parsing of first argument (Bug#28323)

- #30846 26.0.91; debug-watch of kill-all-local-variables triggers
      'assertion failed: found == !EQ (blv->defcell, blv->valcell)'
  Bug in new feature.
  - Fix bug#30846, along with misc cleanups found along the way