Reland "[regexp] Include unicode/uvernum.h in parser"

This is a reland of 0db90bc
Original change's description:
> [regexp] Include unicode/uvernum.h in parser
>
> This patch explicitly includes unicode/uvernum.h in the regular
> expression parser.
>
> It should be removed once we no longer need to check
> `U_ICU_VERSION_MAJOR_NUM` during preprocessing, i.e. once Node.js
> updates their ICU. This is an ongoing effort:
> nodejs/node#16876
>
> BUG=v8:4743
>
> Change-Id: I3cd9447b481249a9035d9fb00745057da8809c58
> Reviewed-on: https://chromium-review.googlesource.com/758407
> Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Commit-Queue: Mathias Bynens <mathias@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49253}

Bug: v8:4743
Change-Id: Id3f375f27fb5eaa4129884f99095d16763bd6e86
Reviewed-on: https://chromium-review.googlesource.com/758861
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Mathias Bynens <mathias@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49260}

src/regexp/regexp-parser.cc

@@ -16,6 +16,9 @@
 
 #ifdef V8_INTL_SUPPORT
 #include "unicode/uniset.h"
+// TODO(mathias): Remove this when we no longer need to check
+// `U_ICU_VERSION_MAJOR_NUM`.
+#include "unicode/uvernum.h"
 #endif  // V8_INTL_SUPPORT
 
 namespace v8 {