Test: Cover mt_mbuf_time_stamp RX offload dynfield path

[DawidWesierski4]

What

Adds a unit gtest covering the generic RTE_ETH_RX_OFFLOAD_TIMESTAMP mbuf
dynfield path in mt_mbuf_time_stamp(), independent of ice's
IEEE1588-specific packet_type/ol_flag markers (mb->timesync,
RTE_MBUF_F_RX_IEEE1588_PTP/_TMST).

Why

While reviewing an upstream DPDK net/ice patch that removes the
packet_type == RTE_PTYPE_L2_ETHER_TIMESYNC gate around ice's
IEEE1588-specific RX timestamp path (because ice's ptype table never
actually produces that value, making the gate dead code), it became
useful to confirm MTL doesn't strictly depend on that ice-specific
mechanism for RX hardware timestamps: mt_mbuf_time_stamp() already
has a separate, unconditional path via the generic
RTE_ETH_RX_OFFLOAD_TIMESTAMP mbuf dynfield.

Two cases:

• CorrectWithoutIeee1588Markers — with MT_IF_FEATURE_RX_OFFLOAD_TIMESTAMP
  set, mt_mbuf_time_stamp() returns the injected dynfield timestamp
  exactly, even when the mbuf carries none of the ice IEEE1588 markers
  (i.e. the state produced by unpatched upstream ice).
• ZeroWithoutOffloadTimestampFeature — with the feature flag absent,
  the same dynfield value is ignored (falls back to
  mtl_ptp_read_time()), proving the first case isn't vacuous.

No lib/ changes — this is test-only, new coverage of existing,
already-correct behavior.

Testing

meson setup build_unit -Denable_unit_tests=true
ninja -C build_unit
./build_unit/tests/unit/UnitTest --gtest_filter='PtpDynfieldTimestamp.*'

Both cases pass across repeated runs.