Remove hook_page_build() and hook_page_alter()

Existing CR https://www.drupal.org/node/2357755 updated.

(Reported by berdir, discussed with catch, who really wanted the BC layer originally, but now agrees that we can remove it.)

#6: I intentionally didn't remove those mentions because it's not clear what to replace them with; we already removed "D7-style hook_page_build()", where you can add stuff to the page; HEAD's hook_page_build() can only add attachments. #2352155: Remove HtmlFragment/HtmlPage will remove DefaultHtmlFragmentRenderer and DefaultHtmlPageRenderer and hence fix this for us automatically. However, I understand your request, and I've complied with it as well as possible, given the transitional state we're in.

#7: You're right, test coverage for this is incomplete, which is why the IS says that the BC hook_page_build() and hook_page_alter() are broken. In the discussion until #5 (and in IRC), we've decided to remove hook_page_build() and hook_page_alter(), because they were misleading anyway.
Removing them also fixes the problem of them breaking hook_page_attachments() and hook_page_attachments_alter(): if there were any hook_page_build() or hook_page_alter() implementations, they would override (!!!) what all hook_page_attachments() and hook_page_attachments_alter() had added before and including in system.module's implementation.
Clarified that in the IS.

My apologies to all of you, I really dropped the ball on this one. Clearly I should have added more test coverage — even when like in this case I should've been able to extend existing test coverage, but there wasn't any. That's no excuse. Sorry!

No, we never had tests for attaching assets in hook_page_build(). I did not remove any tests. Look at the patch in #2350949: Add hook_page_attachments(_alter)() and deprecate hook_page_build/alter().

Tests now added.

I'd reroll this, but the remark in #15 is addressed by #2352155: Remove HtmlFragment/HtmlPage already, which is already RTBC. Trying to fix this here would be wasting time. Especially because if this'd land first, I'd have to reroll the other patch, which is 20 times bigger.

This will need to be rerolled once #2352155 lands.

Thank you very much! Just one problem:

+++ b/core/modules/system/src/Tests/Common/PageRenderTest.php
@@ -33,24 +33,6 @@ function testHookPageAlter() {
@@ -61,7 +43,14 @@ function testHookPageAttachmentsAlter() {

@@ -61,7 +43,14 @@ function testHookPageAttachmentsAlter() {
   function assertPageRenderHookExceptions($module, $hook) {
     // Assert a valid hook implementation doesn't trigger an exception.
     $page = [];
-    drupal_prepare_page($page);
+    $page = drupal_prepare_page($page);
+    $expected = [
+      'library' => [
+        0 => 'core/foo',
+        2 => 'core/baz',
+      ],
+    ];
+    $this->assertEqual($page['#attached'], $expected);
 
     // Assert an invalid hook implementation doesn't trigger an exception.
     \Drupal::state()->set($module . '.' . $hook . '.descendant_attached', TRUE);

You're missing this hunk from #12.

Haha, oops! And I'm the one who landed that patch! :P I just diffed my last patch and the one rerolled here, and that was the only hunk missing. I should've inspected that hunk more closely.

Reuploading the patch in #20. Thanks again, Pinolo! :)

RTBC per #14, Pinolo just rerolled it after #2352155: Remove HtmlFragment/HtmlPage landed and I just reuploaded his patch after my criticism was wrong.

#31: the CR was already updated in #2.

#34: HAH! Fixed.