Fix unrepresentable float value in ScalarTest

Summary: float can't represent the given value in the literal, so we get this UB error: `runtime error: 1.23457e+48 is outside the range of representable values of type 'float'`. The test seems to not rely on this specific value, so let's just choose a smaller one that can be represented.

Reviewers: uweigand

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D42338

llvm-svn: 323081