Fix MCMC.maxchanges = Inf integer coercion warning in tergm

[smjenness]

Replace MCMC.maxchanges = Inf with .Machine$integer.max in netdx(), control.net(), and netest() defaults.

When Inf is passed as MCMC.maxchanges, it flows to tergm::tergm_MCMC_slave() which calls as.integer(maxchanges) before the .Call to MCMCDyn_wrapper. as.integer(Inf) produces NA with a warning: "NAs introduced by coercion to integer range".

Note that tergm wraps the analogous MCMC.maxedges parameter with deInf(maxedges, "maxint") before as.integer(), which correctly converts Inf to .Machine$integer.max. The same deInf() wrapper is missing for MCMC.maxchanges in tergm_MCMC_slave(). This is arguably a bug in tergm -- see the relevant code:

tergm::tergm_MCMC_slave:
maxedges <- NVL(control$MCMC.maxedges, Inf)
maxchanges <- control$MCMC.maxchanges
z <- .Call("MCMCDyn_wrapper", ...,
as.integer(deInf(maxedges, "maxint")), # <- wrapped
as.integer(maxchanges), # <- NOT wrapped
...)

This warning surfaced after ergm 4.12.0 (2026-02-17) triggered a recompilation of the tergm binary, likely changing how the C code uses the maxchanges argument.

Using .Machine$integer.max (2,147,483,647) is functionally equivalent to Inf for this purpose -- it effectively removes the limit on MCMC changes per step -- while avoiding the integer coercion warning.

• Related to MCMC.maxchanges = Inf #920
• Issue for tergm filed at Missing deInf() wrapper for MCMC.maxchanges in tergm_MCMC_slave causes integer coercion warning statnet/tergm#136

[smjenness]

@AdrienLeGuillou : could you review this before I merge it in? Although you made the Inf change way back in 2024, I think some new warnings were triggered by recent updates to ergm last week.

[AdrienLeGuillou]

I agree with your assessment.
I will check next week with HIV-p if things need to be updated as well there for this and the deprecated ERGM term (similar to #953)