mojo: Remove StringTraits::IsValidUTF8.

This should only affect the Mojo.InvalidUTF8String histogram, since
nothing else appears to rely on this. I would expect that histogram
to change (in one direction or another, depending on which of the
following is more common in the wild):

The WTF implementation of this trait was incorrect, for a few reasons:

* WTF::String::FromUTF8 returns a null string if the input was invalid
  UTF-8, but StringTraits<WTF::String>::IsValidUTF8 returns true in this
  case.

* Plenty of valid UTF-8 strings get a 16-bit representation in Blink,
  but StringTraits<WTF::String>::IsValidUTF8 returns false in this case.
  For example, any string with a CJK or Emoji character would be
  erroneously reported as invalid.

* Plenty of valid UTF-8 strings get an 8-bit representation in Blink
  which is valid Latin-1, but not valid UTF-8. For example, strings
  with non-ASCII characters (e.g., accented Latin characters) may be
  erroneously reported as invalid, depending on what character follows.

Instead, this should check that the input from which the string was
constructed was valid UTF-8. This check is the same regardless of what
the destination type is, so it's moved directly into Deserialize.

Bug: 699319
Change-Id: I078830fde108650d570c90e6248bd5255c89ab2c
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/3999828
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Commit-Queue: Jeremy Roman <jbroman@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1066711}