Fix set max_rd_atomic respecting the local device limit

[sfff]

When configuring max_rd_atomic account for our device's constraint in cases where the destination's max_qp_rd_atom (from ibv_devinfo -v) exceeds our local limit.

[sshaulnv]

@sfff, Thanks for the PR.
The change looks great, except the fact that the it will print report with the original outstanding reads.
please fix the report print and we can have it merged.

[sfff]

@sshaulnv, Thank you for the review and the improvement suggestion. I agree that the current approach may appear somewhat complex.
That said, the existing output scheme, although not very simple, already provides all the required information:

1. The test header displays the current parameters print user_param->out_reads
2. After exchanging information with the remote side, the output includes details about Outstand reads on both the local (local address: LID 0000 QPN ... OUT ...) and remote sides (remote address: LID 0000 QPN ... OUT ...) print remote OUT

Would it make sense to rename "OUT %#04x" to "OUTSTAND %#04x"? Alternatively, is there a better way to improve the clarity of the output without breaking the established convention?

I would like to additionally point out that this patch is intended to prevent an error in cases where the local value (printed in test header) is lower than the value obtained from the remote side.