Fix calling default constructor to use uniform initialization.

[sukeya]

Hello,

I faced an error that let C be a template parameter, C() is interpreted as a function call when C is a type alias. So, I replaced all ambiguous expression C() to uniform initialization C{}. Could you please merge this PR?

[cary-ilm]

This looks fine, thanks. For completeness, could you add a test somewhere (src/ImathTest/testVec.cpp maybe?) that fails with the current () initializers but succeeds with the change?

[sukeya]

I added a test case which fails with the current () initializers, but it requires CUDA.

[cary-ilm]

We discussed this in today's technical steering committee meeting, and we support this contribution in principle and we'd like to work towards accepting it.

We'd like to confirm what versions of CUDA this relies on, and confirm that the #ifdef's properly exclude ones it doesn't.

We'd also like to add a CUDA validation job to the CI.

We are planning a v3.2 release in mid-August, which will be cut from the main branch, so we'd like to wait until after that before merging the PR. That will provide more time for evaluation before official release.

In spite of my earlier advice to add a test, your original submission to switch to uniform initialization is fairly uncontroversial, so if you'd like to submit that as a separate PR, we can accept that right away, while we work through the issues with the addition of the CUDA test.

Again, thanks for your contribution and your patience.