Options cache updated by update_option() and add_option() even on failure

[jdgrimes]

The add_option() and update_option() functions both update the 'alloptions' cache before performing the insert/update query on the database. It is therefore possible to do the following:

add_option( 'test_option', 'foo' ); // Lets say this fails.
$option = get_option( 'test_option' ); // $option == 'foo' !

But now if we attempt to delete or update the option, we can't:

var_dump( delete_option( 'test_option' ) ); // bool(false)
var_dump( update_option( 'test_option', 'bar' ) ); // bool(false)

This was introduced in [4855].

[jdgrimes]

This does not only affect the 'alloptions' pseudo-option, it affects any option.

[jdgrimes]

First pass at this. Moves the cache updating to inside the if ( $result ) blocks.

[jdgrimes]

That patch moves the cache updating to inside of the if-success blocks for these two functions. I'm not sure if that is the best approach. Should we be invalidating the cache in some way if the add/update queries fail, or should we assume that the cache is still reliable? This patch does the latter.

[nacin]

The patch on #24933 handles this. Please review! Let's leave this ticket open, though, and close them out together.

[nacin]

In 25664:

When queries fail in option functions, bail before setting cache.

Standardize variables so things are less confusing.

fixes #25015.