Update brian2cuda to use current brian2 master

[denisalevi]

There seem to be some changes in the spikequeue in brian2.

Following PRs from brian2 referenced possible changes for brian2cuda:

• [MRG] Faster before run brian2#703
• Remove hard-coded assumption that summed variable is post-synaptic brian2#705

[denisalevi]

The synapse generation has changed (brian-team/brian2#600). But since this is happening on host side in brian2cuda currently, we should be able to use the algorithm from the corresponding cpp template.

[denisalevi]

@moritzaugustin Are we updating to brian2 2.0 as new frozen or are we trying to stay with current master? Brian2genn depends on brian2>=2.0, so this looks like they keep following brian2 master, which is also stated here (brian-team/brian2genn#25)

[moritzaugustin]

let's have brian2.0 as new frozen commit which might seem unnecessary at the moment but we want to follow brian2 master later if incompatible changes happen (let's discuss then again whether to have 2 branches in this case)

[denisalevi]

We are now following brian2 commit fadc6a0aeb90d1b4d343470628457d8561536f67. There are still a few tests failing from the brian2 test suite, but they are not of high priority right now. I have opened issues for those (#46, #47, #48, #49, #50, #52 ).
Closing this issue.