Develop merge master

See merge request !25

See merge request !24

Fixes for XCI parsing in Python 3.6

See merge request !23

- Support JSON format XCI files (#119)
- Support XCIX files (#120)

Rework XCI file parsing

See merge request !22

- Support JSON format XCI files (#119)
- Support XCIX files (#120)

Add partial support for VHDL generic packages

See merge request !21

This adds support for generic packages declaration dependency
tracking. It is not complete though, it only works when instantiating
a generic package for using withing the same file.

so hdlmake can satisfy dependencies declared by qips

Following commmit implements to standard makefile argument
"syn_properties" as well additional 3:

- syn_instances -> set_global_instance
- syn_location_assignments-> set_location_assignment
- syn_instance_assignments -> set_instance_assignment

The original foreseen only

- syn_properties -> set_global_assignment

Following brings possibility to specify in manifest in more detail for
example logilock regions.

In order to make the parameters compatible with quartus manifest
properties 'to' and 'region_id' can be now preprended by tilde (~) to
announce that the given value of the command shall be enclosed in
quotes. This is needed as it differs from parameter to
parameter. Example:

following does not need any quotes:

set_global_assignment -name IOBANK_VCCIO 3.3V -section_id 3A

but following does:

set_instance_assignment -name LL_MEMBER_OF "bst_sync:i_bst_sync" -to
"VfcHdApplication:i_VfcHdApplication|bran:i_bran|bst_sync:i_bst_sync"
-section_id "bst_sync:i_bst_sync"

Needed to setup certain global properties, as e.g.

set_global_assignment -name OUTPUT_IO_TIMING_FAR_END_VMEAS "HALF SIGNAL SWING" -rise

lib components are ignored during dependency search

in addition to original CLEAN_TARGETS this commits as well:

- a5_pin_model_dump.txt
- *.sld

to be erased when 'make clean' is issued

Fix lattice diamond take2: Filter clean targets - Add only relevant stages for every tool

See merge request !18

Fix(?): Use context manager for writing Makefile, improve TestSuite on Windows

See merge request !19

These testcases are intended to provoke HDLMake
Errors, therefore not generating a synthesis Makefile.

Before making the Makefile write 'atomic' with
contextmanager, these Makefiles have been generated both
on Linux and Windows.
On Linux, the testsuite passed even then.
But on Windows, these Makefiles have been reported
as busy while the TestCase tried to delete the files.
This resulted in OsErrors.

TestSuite failing on Linux:
FAILED test_all.py::test_quartus035 - FileNotFoundError: [Errno 2] No such file or directory: '035quartus_err/Makefile'
FAILED test_all.py::test_quartus036 - FileNotFoundError: [Errno 2] No such file or directory: '036quartus_err/Makefile'
FAILED test_all.py::test_quartus037 - FileNotFoundError: [Errno 2] No such file or directory: '037quartus_err/Makefile'
FAILED test_all.py::test_quartus038 - FileNotFoundError: [Errno 2] No such file or directory: '038quartus_err/Makefile'

Probably this is because these testases SHALL NOT generate a valid
makefile.
Before 'atomizing' the Makefile write, these Makefiles have been
generated anyway, despite the detected and reported errors.

No funstional change. Not fully clean refactor.
TestSuite Still passing on Linux.