This is to follow the naming convention and easy coding, if you know that a
signal is a single-cycle pulse, you don't need to worry about edge detection

This ext_pll_10_to_125m component is part of xilinx platform-dependent
folder. It is used in spec_top entities. Since it was not in the package,
it had to be declared in top files which is not perfect. So, its
declaration was added to the pkg.

it was hard to get fully coherent statistics, especially if they were changing
fast (e.g. 250kHz in btrain). Setting snapshot bit to high copies at the same
instant all the counters to separate registers. as long as snapshot bit is high,
these registers are exposed to WB and SNMP. so one can read it as long as it
takes and the data is still coherent.

this can be useful when reseting remotely and wanting to make statistics. in
such case the SNMP server
1. reads the reset timestamp
2. sets reset high
3. reads the timestamp of the reset
4. sets rest low, new timestamp is done to have exactly the time of start of
   stats

added info about reset to the output registers, as well as the value of the
input register. fixed stats to work with remote reset

The statistics collected by WR Streamers are passed to WRPC through generic-size
array of 32-bit std_logic_vectors. Such communication is intended to accommodate
communication of WRPC with any user-module that needs to be interface via SNMP

the old rx_lost_o indicates as before two "problems"
1. one or more frames have been lost
2. one or more blocks inside the frame have been lost (probably because
   of CRC error)
However, now we have 3 more outputs:
rx_lost_blocks_o     - indicates only losses of block(s)
rx_lost_frames_o     - indicates only losses of frame(s)
rx_lost_frames_cnt_o - when rx_lost_frames_o=HIGH, indicates how many
                       frames have been lost

changelog:
- changed layout and sequence numbering:
  * layout: ethernet header | timestamp | seq id | block 0 + CRC | 0xcafe + inter-frame seq_no| block 1...
  * the seq_no is per-frame, internally, there is internal sequencing, first
    block is considered 0 but has only seq_no, not sub-frame seq_no, the latter
    blocks have sub-frame seq_no
- added legacy generics
- debugged

The added code is based on the one provided for White Rabbit Core Hands-on
Training (http://www.ohwr.org/projects/wr-cores/wiki/Handson_training). The
code has been only adapted to
- fit the structure of wr-cores repository
- work with the newest wrpc-v3.0

Conflicts:
	ip_cores/general-cores

kintex-7 reference design updated such that it can do grand-master mode (putting the DIO card on the CLB FMC connector and inputting 10 MHz and PPS).

built from commit 5eeb579d in wrpc-sw repo

Otherwise we do twice the same write to memory and we permanently lose one
nrx_avail. This way, after some time minic may drop frames because it will think
its buffer has not enough space (based on nrx_avail value).