improvements to --warn-long test flag

[orlitzky]

As a follow-up to #32973 and #32981, there are a few other ways we could improve the --warn-long flag to keep our test suite speedy:

1. Add a flag like -Werror to the test runner, so that warnings are turned into errors that are reflected in the exit code. This will allow the CI to catch them. Another idea would be to have an --error-long flag that triggers errors instead of warnings at an ostensibly higher limit.
2. Summarize the --warn-long warnings at the end of the test run, so that they're all visible to the user without having to dig through a log file.
3. Add a separate option for warning about # long time tests. Currently when running with --long, you have to use a very long warning time, which then essentially hides the warnings for non-long tests that take too long.

CC: @tornaria

Component: doctest framework

Issue created by migration from https://trac.sagemath.org/ticket/32995

[orlitzky]

Description changed:

--- 
+++ 
@@ -2,5 +2,5 @@
 
 1. Add a flag like `-Werror` to the test runner, so that warnings are turned into errors that are reflected in the exit code. This will allow the CI to catch them.
 2. Summarize the `--warn-long` warnings at the end of the test run, so that they're all visible to the user without having to dig through a log file.
+3. Add a separate option for warning about `# long time` tests. Currently when running with `--long`, you have to use a very long warning time, which then essentially hides the warnings for non-long tests that take too long.
 
-

[orlitzky]

Description changed:

--- 
+++ 
@@ -1,6 +1,6 @@
 As a follow-up to #32973 and #32981, there are a few other ways we could improve the `--warn-long` flag to keep our test suite speedy:
 
-1. Add a flag like `-Werror` to the test runner, so that warnings are turned into errors that are reflected in the exit code. This will allow the CI to catch them.
+1. Add a flag like `-Werror` to the test runner, so that warnings are turned into errors that are reflected in the exit code. This will allow the CI to catch them. Another idea would be to have an `--error-long` flag that triggers errors instead of warnings at an ostensibly higher limit.
 2. Summarize the `--warn-long` warnings at the end of the test run, so that they're all visible to the user without having to dig through a log file.
 3. Add a separate option for warning about `# long time` tests. Currently when running with `--long`, you have to use a very long warning time, which then essentially hides the warnings for non-long tests that take too long.