{edit,fault}.c: improve SIGWINCH and SIGINT handling

The fault.c and edit.c changes in this commit were inspired by
changes in the 93v- beta but take a slightly different approach:
mainly, the code to update $COLUMNS and $LINES is put in its own
function instead of duplicated in sh_chktrap() and ed_setup().

src/cmd/ksh93/sh/fault.c:
- Move code to update $COLUMNS and $LINES to a new
  sh_update_columns_lines() function so it can be reused.
- Fix compile error on systems without SIGWINCH.

src/cmd/ksh93/edit/edit.c:
- ed_setup(): Call sh_update_columns_lines() instead of issuing
  SIGWINCH to self.
- Change two sh_fault(SIGINT) calls to issuing the signal to the
  current process using kill(2). sh_fault() is now never called
  directly (as in the 93v- beta).

src/cmd/ksh93/sh/main.c: sh_main():
- On non-interactive, call sh_update_columns_lines() and set the
  signal handler for SIGWINCH to sh_fault(), causing $COLUMNS and
  $LINES to be kept up to date when the terminal window is resized
  (this is handled elsewhere for interactive shells). This change
  makes ksh act like mksh, bash and zsh. (Previously, ksh required
  setting a dummy SIGWINCH trap to get auto-updated $COLUMNS and
  $LINES in scripts, as this set the SIGWINCH signal handler to
  sh_fault(). This persisted even after unsetting the trap again,
  so that was inconsistent behaviour.)

src/cmd/ksh93/include/shell.h:
- Don't define sh.winch on systems without SIGWINCH.

src/cmd/ksh93/sh.1:
- Update and tweak the COLUMNS and LINES variable documentation.
- Move them up to the section of variables that are set by the
  shell (which AT&T should have done before).

NEWS

@@ -3,6 +3,12 @@ For full details, see the git log at: https://github.com/ksh93/ksh/tree/1.0
 
 Any uppercase BUG_* names are modernish shell bug IDs.
 
+2022-07-01:
+
+- In scripts, $COLUMNS and $LINES are now kept up to date in scripts at
+  initialization and when the window size changes (previously, this
+  required setting a dummy trap for the SIGWINCH signal in the script).
+
 2022-06-28:
 
 - Fixed a bug that caused the <#((num)) or >#((num)) arithmetic seek