Defer register inline script in WP_Widget

[spacedmonkey]

In the widget classes, WP_Widget_Custom_HTML and WP_Widget_Text, the method _register_one is called. This method registered a inline script like this.

wp_add_inline_script( 'text-widgets', sprintf( 'wp.textWidgets.idBases.push( %s );', wp_json_encode( $this->id_base ) ) );

wp_add_inline_script( 'custom-html-widgets', sprintf( 'wp.customHtmlWidgets.idBases.push( %s );', wp_json_encode( $this->id_base ) ) );

This results in a script registering. However, this inline script is not used, unless script is loaded in admin_print_scripts-widgets.php hook.

This inline script being registered like this, means that when the widget is registered, this inline script is registered as well. In my use page types, this inline script serves no purpose and hiders performance.

[spacedmonkey]

Before, see red square

[spacedmonkey]

After

Trac ticket: https://core.trac.wordpress.org/ticket/57864

[sakibmd]

PR looks good @spacedmonkey and thanks for figuring it out.

@felixarntz Actioned feedback and these and seeing performance improves still.

[spacedmonkey]

In 55553:

Widgets: Defer register inline script in WP_Widget_Custom_HTML and WP_Widget_Text.

In [41376] an inline script was added to push idBases for the custom html and text widgets. However, this script is not used unless the widget script is output in the widget screen / customizer. Deferring registering this script until it is needed, results in a faster server response times.

Props spacedmonkey, sakibmd, flixos90, westonruter.
Fixes #57864.

[spacedmonkey]

In 55554:

Build/Test Tools: Revert [55553].

Revert package-lock.json changes in [55553].

See #57864.

Committed

[flixos90]

As part of investigating #57916, I found that this enhancement is particularly beneficial for widgets_init performance, and it also improves overall performance slightly, which is great to see. Details in https://core.trac.wordpress.org/ticket/57916#comment:12.

I don't think it should be backported since it doesn't fix any 6.2 regression, but rather is a new performance enhancement, and the win is also not large enough to justify backporting. But it's great that you caught this, it is indeed somehow indirectly contributing to the WP 6.2 slowdown of widgets_init.