Awgtpman GDS reconciliation, new build

See merge request cds/advligorts!177

contains the rate at which the IOP is sending data to user application
models. This was added to support awgtpman when IOP is passing data at
other than normal 65536 samples/sec.

awgtpman: removed relative paths from some target_include_directory() calls.  Added some missing headerfiles that should have been in the last commit

GDS/AWGTPMAN: missing header files included in src/gds/awgtpman/dtt.  Also gmutex files added to src/gds/base.  These are used by NDS and are separate from awgtpman. Now builds in a pristine container with only mk-build-deps isntalled.

Awgtpman reconciled with GDS replaces previous awgptman.  Build changed to DTT style awg build.  The main difference is that rpc files are built out of tree.

A GDS linkage to daqd has been removed.  This will need to be addressed in a future commit.

checkdaqconfig: filling another hole in the master file parsing.

See merge request cds/advligorts!213

checkdaqconfig: filling another whole in the master file parsing that would cause an exception.  Also, updated the test file to work with the latest version of pytest.

Daqd update, serve data from circular buffers before going to disk if possible.

See merge request cds/advligorts!212

Consolidated the online and in memory logic in the net_writer into one location, making sure the absolute gps request gets a termination block.

Updated the nds live test to also test doing a in buffer test with absolute gps times.

Previously requests that did not have a start/delta time could go from the buffer by default, and a start/delta would read from frame first and then finish up in reading from ram.
The implications are that short reads that use absolute gps times that fall in the time span the data is available but where a frame has not been written yet will fail as it will attempt
to read a frame, fail, then return.  This pushes a check up front to see if the start/delta is caught completely in the in memory buffer and then attempts to read that before falling back to reading from frames.

 * Changes needed to make passed in parameters const char* instead of char*.
   * This included cleaning up a number of other functions that should have been able to work with const char*
 * extracted some network address and parsing functions from the daqd and net_writer class into standalone functions.
   * This was done to allow unit testing as it otherwise pulled in all of daqd.
 * In the start * writer code, replaced the gotos with returns as we are now in a standalone function.

clocking.

error in state word to use alarm high setting of 23 usecs.

on board and setup accordingly. This done because LIGO has both
8 and 16 channel versions of this card.

Fix timer in user IOP

See merge request cds/advligorts!210

- Change ADC timeout in iop_adc_functions
- Remove Demand DMA mode set to BCR: Not correct for this ADC model

Refactoring of the daqd/comm.y code to move the start (net/file/...) writer...

See merge request cds/advligorts!211

We no longer do NO_BROADCAST and always to testpoints.

Ran clang-tidy to add braces to all the if statements.

 * Done in two steps, 1st extract the code w/ minimal reformatting
 * Clean up formatting

Refactoring of the daqd/comm.y code to move the start (net/file/...) writer code out of the bison/yacc file.

Move the code out and replace the positional arguments with named variables with known types to simplify code introspection/debugging/...

This matches the behavior of the userspace IOP to the kernel IOP.
If the gps second is not incremented at the start of cycle 0,
it causes IPC errors as receiver and sender timestamps disagree.

The previous version did not handle overflow correctly,
which resulted in large excursions in the cycle time.

and GSC16AO16 modules.