Some more thoughts ..
I figure out you need more rudder to keep it during turn and if you already were close to Vmc you are close to rudder stop, so, is it possible to do these turns without losing control?
There'll be a degree of Type specific behaviour involved.
if I'm flying 5 kts above Vmc and banking (i.e. moderate bank 30 degrees, no need to do aerobatics on one engine 
)
Suggest that is not a good idea. Until you are at normal OEI speeds, bank should be constrained significantly. 10-15° is a suitable ballpark. Too much risk of letting things get away from you at large bank angles.
even if reaching Vmc, there was still some aileron control to return to wings level or 5 deg toward the live engine and so regaining straight flight,
Should generally be the case. The main problem is that, in the few knots near Vmc, things may happen very quickly and, unless the pilot is right on the ball, the risk of departure is high. Back near Vmc, the pilot's immediate actions in the event of failure must include a rapid introduction of some favourable bank.
if Vmc=loss of control at all, i.e. even roll loss of control, if so, every takeoff would be fatal when you lose one engine.
Vmc doesn't necessarily infer loss of control. Rather a competent pilot should be able to control the aircraft within prescribed parameters. Further, it is usual to have a modest margin above Vmc for the min speed takeoff speed schedule. What this all means is
(a) there is NO time to play Monday morning quarterback in the cockpit. Actions need to be well rehearsed and very prompt.
(b) at the very light weights required for min speed schedules, aeroplane acceleration is very high during the failure rundown
(c) the need is for rapid and aggressive pilot control inputs to prevent the situation getting to a point where things do become irrecoverable.
(c) the aeroplane at higher speed schedules (higher weight or overspeed) is a comparative pussycat. At min speed, aft CG, etc., the pussycat very rapidly becomes a raging tiger ..
One of my worries is the case where pilots are not trained in min speed schedule failures, routinely fly (and train) at overspeed, and then (occasionally) are required to fly from airports requiring min speed protocols. As I have seen in the sim environment .. an eye-opening occurrence if the failure is introduced for the latter circumstance. Comparatively easy to train out the problems but, if that is not done, the first real life event, potentially, may be catastrophic due to startle factor time delays.
you lose just the ability to maintain a straight track, i.e. rwy centerline during takeoff, but not the ability to roll the aircraft.
Things are not quite that bad.
Failure will involve yaw .. which can be controlled. Lower skill levels will see drift .. which can be controlled with bank. A modicum of roll control should remain. The pivotal consideration is skill level and that comes down to training to task. There is no room for figuring out what one needs to do at the time .. the responses must be automatic, instinctive, and rapid.
It is not overly difficult,
with training, to have a min speed schedule V1 failure and track out accurately.
For instance, my training technique for this was to have the student progressively work up to a nil, or near-nil, vis takeoff, min weight, min speed, max aft CG, worst failure the box had programmed, then have the failure shortly after VR (ie during the rotation flare with nil vis reference), and track out on the LLZ. A few practice runs and, with aggressive initial control inputs, it would work out reasonably well. Great confidence builder when the student then could recover via a hand flown, raw data, landing in 0/0 conditions. The environment would get a tad sweaty but it was all in good humour. Of course, this all required a VERY good standard of I/F scan and manipulation ... that was eminently trainable in the sim, even for 200 hour cadet folks.
So how to bank 5 degrees towards live engine to assume best attitude for Vmca if you don't have enough aileron authority?
There should be adequate roll control left to do that. However, the essence is time .. back at (or close to) Vmc, the pilot just does not have the luxury of scratching his beard for a bit while contemplating the situation during a failure .. the response must be automatic.
Yes, but I thought you are reached at full rudder considering you are below (actual) Vmc (which is increased due to the fact that you are banked towards the dead engine.)
Keep in mind that bank is largely for static considerations and the effect of slip angles. You are having some difficulty separating static and dynamic matters as well as time lags for inertial things to start happening.
The Air North mishap thread is worth a read. More importantly, the
BASI animation in the report is very sobering .. (click on the AVI hyperlink under
"Related Files".)