reflect: Use `float64` string representation for `float32` reflection equivalency

[SBGoods]

Relates: #1014

This fixes a bug encountered when using function.Float32Parameter in a provider defined function, similar to issue #914:

=== RUN   TestFloat32Function_known
    float32_function_test.go:20: Step 1/1 error: Error running pre-apply plan: exit status 1

        Error: Error in function call

          on terraform_plugin_test.tf line 13, in output "test":
          13:                   value = provider::framework::float32(1.23)
            ├────────────────
            │ while calling provider::framework::float32(float32_param)

        Call to function "provider::framework::float32" failed: Value Conversion
        Error: An unexpected error was encountered trying to convert to number. This
        is always an error in the provider. Please report the following to the
        provider developer:

        cannot store 1.23 in float32.
--- FAIL: TestFloat32Function_known (0.36s)

The current reflection logic for float32 compares the reflection target's string representation of the float32 value with the string representation of that same value in a new big.Float object and throws an error if the representations differ. Creating a new big.Float object requires a conversion from float32 to float64 resulting in precision loss and differing string representations.

This PR switches the string comparisons to float64 and adds overflow and underflow checks for float32.