The point is: If you have a (hollow) device which desnity is belwo that of water it will flow.
so answer to the trick question: If you make a very thin foil from the stell, say about 1/1000 mm (which is still feasibly) and then make a hollow cylinder from that the length of the cylinder needs probably be no more than another 1/1000 of a millimeter to make the devices' density below 1. so 2/1000 mm deep water would suffice...
The answer to the second question goes along the same line, but we assume that you meant a closed device. So we take a hollow sphere made from 800 kg steel, filled with air. So that the density is 1 kg/l. The 800 kg steel have a volume of 100 l. Every l of air is lifting 1 kg steel (roughly air has a density of 1 g/l, water of 1 kg/l, the error is 0.1 % ). So we need a sphere with a hollow volume of 700 (or correctly 701) l. This inner (hollow) sphere has a radius of about 5.51 dm (the rounding error is bigger than the error I introduced earlier).
The outer radius of the hollow speher is (going from 800l): 5.76 dm, so the wall thickness is approximatly 0.25 dm or 2.5 cm or roughly an inch.
Nice work. And yes, for a closed surface, a sphere would be no doubt ideal. If you would accept a shape that is open at the top, such as a boat, with max volume subject to the requirement that there is no linear route from within the boat into the water surrounding it that doesn't pass through some pre-fixed minimum length of iron, the most natural possibility would probably be a hemisphere..?
Floating via the lift due to the liquid displaced by the steel container, the way boats and ships do. When that is taken to extremes, obviously there are other issues to take into account, as you pointed out. Will it even be possible to make water form as a one molecule thick layer, or will surface tension turn the task similar to trying to push a balloon through a narrow crack.
Using oil instead of water could be an option. It's slightly less dense, and thus has less lift, but has no surface tension so it will spread out to a more narrow layer. With oil, what I get is that you'd need 3.6 cubic millimeters of oil coating a a hemisphere-shaped hole in which the hemisphere-shaped steel thingy floats. The steel would need to be about 0.28 mm thick. Feasible of course.. it's in the same ballpark as gold foil. Whether it would retain the hemisphere shape is another matter.