Is a fluid burning boiler more efficient than the fluid burning heat source with a heat exchanger? And what fluid is the most efficient to use? (I did some math and it seems like petroleum gas but idk I'm not good at math)

Petroleum Gas is actually a fairly good one to use.
However, Oil is limited (because it needs specific nodes) where Hydrazine is unlimited as it is made from Air and Water (Which is unlimited in that you just need a lake or something, not specific nodes that reduce in speed) only (split Water into Hydrogen and Oxygen, and Air into Oxygen and Nitrogen, and then its just a complex chain of combining the two)
Pollution is also a factor, Oil products do tend to create more pollution than some alternatives. A combination of different fuels would also work, for example, you're better off converting light oil into Liquid fuel, rather than petroleum gas, ending up in using a combination of Liquid fuel and Petroleum gas.

Fluid burning Boilers (and Fluid burning generators) burn the fuel directly for a 100% efficiency.
The Fluid burning heat sources on the other hand offer Neighbour bonus, similar to a nuclear reactor. On top of this, because they have pass-through pipes, you can actually make a solid grid, so most heat sources can have a bonus on all sides. They do have different bonuses per tier, but if you assume to use the MK3, that has a 100% bonus, meaning you can pull 500% energy from the fuel. (though that drops to 400% on a side heat source and 300% on a corner)
Just keep in mind that you can't mix fluids, so if you do opt to run more than one type of fuel, you'll need a different block of heat sources for each type. (they can still be linked with heat pipes, just not fluid pipes)

So... if you do want a fluid burning power source, the best thing to do is to either burn Petroleum Gas, Liquid fuel and/or Hydrazine in Fluid burning heat sources powering heat exchangers, which in turn power matching tier steam turbines.

Also, in theory, while you can power any heat exchanger from any heat source, or any steam engine/turbine from any steam source, you need to be careful to make sure the heat source can actually get hot enough for the heat exchanger to work (The MK1 heat source can't power a MK3 heat exchanger), and also match the steam temperature for the steam source with the steam engine/turbine (Lower temperature steam will just have a steam engine/turbine underperform, but higher temperature steam will waste energy)

For reference, each fluid's power output, and pollution multiplier:

crude oil = 1.9MJ, 10x
Petroleum Gas = 2.3MJ, 1x
Light oil = 1.5MJ, 2x
Heavy oil = 1MJ, 3x
Liquid Fuel = 2.3MJ, 0.8x
hydrazine = 340kJ, 1.2x
Hydrogen = 45kJ, 0.1x
glycerol = 730kJ, 1x
Deuterium = 3.55MJ, 0.1x

Something seems not quite right about the fluid burning generators. I've got ten of them powering a whole base, which they seem to do on the very bare sniff of fuel. With ten of them (a mix of 3 x Fluid Burning Generator 1 and 7 x Fluid Burning Generator 2), generating about 6MW (so something about 40% of capacity) each has fluid consumption of 0.1/s. Needless to say, they don't get through much fuel. Is that really right?

As far as I can tell, changing scale_fluid_usage to false in prototype/entities/fluid-generator.lua (all three instances) produces the right behaviour - though I'll grant it doesn't look right in the code.

If you set scale_fluid_usage to false, it'll just burn at fluid consumption limit, and if you don't have one of those set, it'll empty it's entire fluid box every single tick, if it generates energy or not (and they do have a power generation limit), so it becomes a fluid void sink.

And, yes, they don't use all that much fuel, especially if you're using a fuel source with a high fuel value.
if you use something like Liquid fuel, or Petroleum gas, then each unit of fuel can generate 2.3MW, which means for 6MW, you'd be using a total of about 2.5 units per second. spread that out to each fuel generator, and it becomes about 0.2/s, yes, it rounds down to the nearist multiple of 0.1/s
if you're mixing MK1s with MK2s, then they'd be consuming them at different rates each, so its very likely the MK1 generators will be displaying 0.1/s

Consider a lump of coal is 4MJ. you'd be burning almost twice as much PG to get the same result. also water becomes 165 degress steam which only has 30kJ per unit, having your steam engine consume 30/s is normal, that's 900kW, PG has more than 75 times the energy in it.

Thanks for the explanation; it does make sense. I'm not finding that setting scale_fluid_usage has quite the effect you describe. I still see the consumption scaled by power output, just nearer 40 or 50 units/s rather than 0.1 or 0.2 units/s.

TBH my priority at the stage of the game I'm at is not to maximise energy efficiency but rather to dispose of excess oil; I'm currently heavily constrained by the availability of sulphur. I'd thought that using the excess for power would be a good way forward, but actually the consumption is so low that it makes little difference.

you can actually just vent petroleum gas. go with the basic recipe that gives sour gas only, make the sulfur gas, and vent the PG.
but there are alternate sources of sulfur too, such as nickel and lead processing.
if you're using bob electronics, you should have a fair use of lead from there, change your lead processing from straight smelting, to the chemical process, and you'll have plenty of sulfur.