Implement qgraf-comptatible keywords for FORM interface

[jodavies]

With this change, the diagram filtering is compatible with qgraf's output.

This passes many tests (to come, for the test suite) of single keywords.

Currently there is a difference, to be resolved, when specifying both of a "dual option". For example, specifying "onepi + onepr" in qgraf results in 0 diagrams. In FORM the options cancel each other, and everything is generated. A change to grcc is required to resolve this.

Also to discuss:

• what should WITHSYMMETRIZE do? GRCC has both GRCC_OPT_SymmInitial and GRCC_OPT_SymmFinal flags

[jodavies]

Updated:

• print an error and terminate, if conflicting options are specified, for eg `onepi_'+`onepr_'
• "external" particles set the `CHECKEXTERN' flag (and it is not available as a pre-processor variable), the result passes some comparisons with qgraf

Still to do:

• should `withsymmetrize_' enable both GRCC_OPT_SymmInitial and GRCC_OPT_SymmFinal? Should we have two FORM keywords to control them separately?
• test floop
• due to the way the options are passed, one has to be careful not to give, eg, `onepi_' + `onepi_' -- I am not sure there is an easy way to avoid this issue without changing how the options work completely.
• `withonepisets_', `withblocks_' "work" in that they change the output, but I don't know if it is correct
• `withedges_' presumably can be easily crosschecked against qgraf

I have validated against qgraf at 0,1,2,3 loops, with all pairs of non-conflicting filtering keywords, for:
Phi34:

• phi - phi
• phi - phi phi
• phi phi - phi phi

Qcd:

• qua - qua
• glu - glu
• gho - gho
• qua - qua glu
• glu - glu glu
• glu glu - glu glu

[coveralls]

coverage: 53.147% (-4.0%) from 57.118%
when pulling 2fc5602 on jodavies:qgraf-keywords
into a405343 on form-dev:master.

[jodavies]

Small update:

• remove `withsymmetrize_', introduce `withsymmetrizei_' and `withsymmetrizef_' to independently control GRCC_OPT_SymmInitial and GRCC_OPT_SymmFinal
• tested: vacuum graphs can be generated at >=2 loops by supplying empty sets for incoming and outgoing, and the number of topologies generated agree with FeynGraph for "PHI3456" at 2-6 loops.
• many of the above stated tests also pass at 4 loops, some take too long and timed out. None which finish, fail.

Qgraf has the dual option "notfloop": we could add this also, with a small change to grcc.cc. There is no reason not to do this, I suppose, just in case someone wants it.

I noticed that generating 6 loop vacuum graphs takes a surprisingly long time. Much longer than 6L propagators, for eg. Below 6 loops there is no obvious difference.

[jodavies]

Update:

• added notfloop_

Now there are open questions regarding exposing GRCC_OPT_Block in the FORM interface and what exactly it corresponds to in Qgraf, and signs assigned when graphs have four-fermion vertices.

But I would say that those questions can be for another PR, and this is now ready to merge so @tueda can add tests to the testsuite using these filtering keywords.