Fix size_of_scalar test

[alamb]

Which issue does this PR close?

N/A

Rationale for this change

This test is important. It verifies that the memory use of code like GroupByHash is not changed. Quoting:

        // Since ScalarValues are used in a non trivial number of places,
        // making it larger means significant more memory consumption
        // per distinct value.

It turns out that the way the cfgs were setup the test never was invoked.

The change seems to have come in via #1455 from @maxburke

I found this while reviewing #2523

What changes are included in this PR?

Fix test so it is always invoked

Are there any user-facing changes?

no

[alamb]

I like this formulation as it ensures that one of these assert_eq! calls will always be invoked, regardless of architecture

[tustvold]

You could also write something like

let expected = match cfg!(target_arch == "aarch64") {
  true => 64,
  false => 48,
};

[alamb]

        let expected = match cfg!(target_arch == "aarch64") {
            true => 64,
            false => 48,
        };

resulted in

error: expected 1 cfg-pattern
   --> datafusion/core/src/scalar.rs:380:30
    |
380 |         let expected = match cfg!(target_arch == "aarch64") {
    |                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |
    = note: this error originates in the macro `cfg` (in Nightly builds, run with -Z macro-backtrace for more info)

for me locally

[tustvold]

Sorry, try removing the double equals...

[tustvold]

A comment about why there is the architectural discrepancy might be nice

[tustvold]

You could also write something like

let expected = match cfg!(target_arch == "aarch64") {
  true => 64,
  false => 48,
};