Merge branch 'feature/ext_weights_refactor' of https://github.com/quetric/finn into quetric-feature/ext_weights_refactor

Adds an end-to-end test for the UNSW-NB15 (cybersecurity/network intrusion detection) MLP. Also includes a minor fix to the PYNQ driver (using correctly-generated tensors with respect to datatype for throughput testing)

includes trained model + Python validation script

Fix incorrect shapes for data packing in Pynq driver

Updates to end-to-end cybersecurity examples

Adjust notebook to changes in bnn-pynq brevitas example

Merge branch 'feature/ext_weights_refactor' of https://github.com/quetric/finn into feature/ext_weights_refactor

[test_move_identical_op_past_join_op]: added test case for new transformation.

* Assign BD address segments based on actual range

* Change minimum address range to 4k

* [Test] lfc end2end tests use runtime-wr weights instead of tfc

Co-authored-by: Yaman Umuroglu <yamanu@xilinx.com>

* [Notebook] tentative PYNQ deployment section

* [Notebook] cybsec-3: workaround for driver bug

* [Deps] update finn-base to get faster binary packing in driver

* [Notebook] cybsec-3: add optional deployment sections

* remove obsolete netron params

* set default ip and port

* set host and port as fixed params

* consistency

* Add MoveMulPastMaxPool transformation to streamlining (#285)

* Add MoveMulPastMaxPool transform to streamlining

* Add second call to transformation within streamlining

* [Docker] set Netron version to >=4.7.9

Co-authored-by: Neil Kim Nielsen <37108154+neilnielsen@users.noreply.github.com>
Co-authored-by: Felix Jentzsch <45395194+fpjentzsch@users.noreply.github.com>
Co-authored-by: Yaman Umuroglu <yamanu@xilinx.com>

* Add MoveMulPastMaxPool transform to streamlining

* Add second call to transformation within streamlining