The most plausible explanation to me is that HAL had to be specifically enabled to control a pod. It was plausible for Poole to enable HAL to do this when taking a spacewalk to replace the AE-35 unit, as he did not expect HAL to do what it did and needed HAL to be able to control the pod to rescue him in case of an emergency or to ease reentering the pod.
Alternatively, as Hypnosifl stated in a comment, it is plausible that Bowman could have easily overriden any attempt by HAL to control the pod while he’s still inside it, so it would not have made sense for HAL to attempt this at that point. As Bowman directly left the pod for the Discovery, the only way for HAL to use the pod afterwards was to damage the Discovery and thus to endanger itself, which it might not have wanted to do.
Even if HAL had full control of the pod, I only see three ways for it to neutralise Bowman:
Ramming the discovery (see above).
Steering the pod away from the Discovery. We do not know on what distance HAL could communicate with the Pod and it is conceivable that this distance is very short. Once communication has ceased, Bowman could regain control and possibly would need to give control back to HAL manually. When Bowman goes to fetch Poole, we briefly see the following screen:
While the displayed information may refer to Poole, it may also refer to the Pod itself, indicating in particular that the pod’s comlink to HAL is broken, which would enable Bowman to freely steer the pod afterwards under the above assumptions.
Keeping the pod in distance until Bowman runs out of air, water or food. If overriding HAL’s control from within the pod was possible, this would have given Bowman plenty of time to do so.
Depressurising the pod (see below).
What Bowman does to rescue himself doesn't seem very intelligent or unexpected, and HAL is supposed to be a very intelligent computer.
When coming to think of it, HAL might not have so many ways at hand to kill or neutralise Bowman that do not allow him to grab a spacesuit and disable HAL.
Actually, the only other way I can think of at all would be to evacuate the ship. However, why would it be able to do this? Yes, it can probably control the airlocks, but I expect them to have independent, mechanical failsafes that avoid a depressurisation of the whole ship.
So, when Bowman heads for the pod and forgets his helmet, it seems not implausible to me that HAL coöperates and lets him go simply because this is the most promising strategy.
Finally, in the movie, HAL may very well be malfunctioning in some way and thus its decisions do not need to be fully logical.