Wrong number of threads used if interactive pool exists.

[IanButterworth]

My fix in #496 introduced a bug where the number of usable threads is overreported by the size of the interactive pool.

NNlib only uses @spawn ... so uses the :default pool.

Using maxthreadid prevents breaking indexing that uses threadid but means too many threads are expected to be available.

Julia 1.12 is being trialed with an interactive thread by default, so it will always be +1 too high on 1.12.

To fix this, I think either:

1. the threadid-indexing needs to be removed
2. go back to nthreads(:default) and make sure threadid-based indexing offsets by the size of the interactive pool

[ToucheSir]

I don't quite understand what either option here would require, but I'm all for simplifying the threading logic for the conv* functions.

[IanButterworth]

threadid indexing was removed from the main package in #497

But I noticed theres some threadIdx stuff in exts. It may not be related.. I'm not sure.

I guess I could PR, include test setups with -t1,1 and see if it breaks.