[BUG] cardinal_t is the worst name

[DenisYaroshevskiy]

For the second time I expect cardinal_t<T> give me expected_cardinal_t<T>
I mean that's quite confusing rly.
Let's delete it.

[jfalcou]

There is an old issue a out that already. We need something that return scalar cardinal for dcalr and correct cardinal or wide anyway.

Remember I alos want to move fixed from type to NTTP so maybe we can have a deep thought on that at this moment.

[jfalcou]

Duplicate of #886

[jfalcou]

After mulling over it, here's a proposal:

• remove scalar_cardinal as it is used mostly nowhere and adapt simd_value accordingly
• remove cardinal_* and use T::size() everywhere it should have been
• rename T::size() to T::lanes()
• rename expected/fundamental_cardinal to something based on lanes

By doing this, it also prepare the change to NTTP lanes in wide type.
What do you think of this ?

[DenisYaroshevskiy]

I do not mind size() or lanes().
This will affect quite some usage links but w/e - we agreed that's what we do.
I don't see why we need scalar cardinal.

[jfalcou]

It was used at one point, then we discard 99% of its use case when we reworked load/store.
We need to rewrite simd_value and it can be eliminated