Rebase/dpu offload

[bosilca]

Port all @yqin changes from the dpu offload branch onto a recent ucx.

[bws]

This fails to build with the following configure options:
../configure --disable-logging --disable-debug --disable-assertions --disable-params-check --with-knem --without-java --enable-devel-headers

Compile error is:
CC base/libuct_ib_la-ib_md.lo
../../../../src/uct/ib/base/ib_md.c: In function ‘uct_ib_md_query’:
../../../../src/uct/ib/base/ib_md.c:259:42: error: ‘UCT_IB_MD_PACKED_RKEY_SIZE’ undeclared (first use in this function); did you mean ‘UCT_IB_MD_PACKED_MKEY_SIZE’?
259 | md_attr->rkey_packed_size = UCT_IB_MD_PACKED_RKEY_SIZE;
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
| UCT_IB_MD_PACKED_MKEY_SIZE