The datasheet listing for Rds(on) does not mention any other voltage than Vgs=10V and the phrase logic-level isn't present.
Delving further it turns out that there are some Rds(on) curves for other Vgs values in the graphs - but the graph section of a datasheet are usually "typical" not "guaranteed" behaviour and should always be treated with a grain of salt. This is the first time I've seen an important fact like a device being logic level buried deep in the datasheet like this - and the performance for the lower Vgs voltage is much worse than at 10V (certainly if you wanted to put 0.2A through it at for full temperature range).
But anyway the device isn't suitable for the intended use on power-dissipation grounds alone - given the abs max power dissipation is 0.2W and the current load 0.15A, any value of Rds(on) greater than about 1 ohm is bad news. Given the guaranteed worst-case Vds(on) is 3.5 ohms, that would mean 0.52W dissipation (even the typical value gives 0.21W which exceeds abs max.)
Also given the spread in Rds(on) at 10V (worst case = 2.5 x typical is a massive variation), the spread of threshold voltages (3:1 ratio), then you have to take those graphs at 2.75V and 4.5V with more than a grain of salt I think - I don't believe the manufacturer is actually claiming it guaranteed to work at 3.3V, just that some specimens are likely to most of the time. Otherwise they would put an entry in the table for Rds(on). Still it does look like it should be upto 5V.