ON-16265: Set no-poison vi flag when ctpio_mode=sf-np

[dchadwic-xilinx]

Since ctpio_mode can be <0, using a signed int is more appropriate. (https://github.com/Xilinx-CNS/tcpdirect/blob/tcpdirect-9.0.0/src/lib/zf/private/stack_alloc.c#L52-L54)

We also decided that there is no obvious reason for the no-poison flag to be set in the event that ctpio_mode < 0. My assumption is that there was an earlier decision that any future 'no-poisoning' CTPIO modes would also have negative values like CTPIO_MODE_SF_NP=-1.

Another solution would be to use an enum but I do not see examples of this being used elsewhere in this codebase. This solution is the least disruptive imo.

[jfeather-amd]

Looks reasonable, have you done any testing? It would be nice to see if you could force CTPIO poisoning with CTPIO_MODE_SF then show that no poisoning happens with CTPIO_MODE_SF_NP although I don't know how easy/plausible that would be to do.

[dchadwic-xilinx]

Testing was done here: proving that the vi_flags are now showing correctly 3000000 rather than 1000000. I'll have a go at forcing poisoning.

[jfeather-amd]

proving that the vi_flags are now showing correctly 3000000 rather than 1000000

This is enough to make me happy, so approving on that basis.

I'll have a go at forcing poisoning

I'd still be interested in the result of this just for my own curiosity, but don't feel that you need to do it - at this point it's more a test of whether the VI flag EF_VI_TX_CTPIO_NO_POISON works!