Optimize dBFS/amplitude conversion functions

[LostRobotMusic]

std::pow(10.f, x) and log10f are painfully slow functions when run per-sample, as is commonly required via dbfsToAmp and ampToDbfs. This PR sacrifices a completely negligible amount of precision for a significant performance boost for several LMMS plugins:

static inline float fastPow10f(float x)
{
	return expf(2.302585092994046f * x);
}

static inline float fastLog10f(float x)
{
	return logf(x) * 0.4342944819032518f;
}

Notice that these are mathematically equivalent operations, not approximations. The lost precision, which is very nearly 0, is solely due to the limited precision of floats.

---

Compiled via GCC with -O2:

1 billion iterations of log10f(x):
31,138,340,305      cycles
8.419502684 seconds time elapsed

1 billion iterations of fastLog10f(x):
15,626,908,607      cycles
4.226729945 seconds time elapsed

fastLog10f is 1.993x as fast on this hardware.

1 billion cycles of pow(10.f, x):
26,823,368,193      cycles
7.258407044 seconds time elapsed

1 billion cycles of fastPow10f(x):
16,135,296,386      cycles
4.315204418 seconds time elapsed

fastPow10f is 1.662x as fast on this hardware.
(This might even provide a larger benefit on older hardware, I've seen performance benefits of up to 3x reported by others.)

---

It might be worth exploring some faster and less-precise alternatives in the future, but what this PR provides is a significant improvement for no actual downsides whatsoever.

[Rossmaxx]

Minor thoughts regarding style.

[messmerd]

Haven't tested it, but I verified that the math is correct [Desmos link], and the changes are minimal, so I think it's safe to merge.

[LostRobotMusic]

@messmerd I get compilation errors when compiling those changes on my PC, std::expf and std::logf are not declared. This might be relevant: https://stackoverflow.com/a/56420862/5759631

The best workaround seems to be simply using expf and logf.

[LostRobotMusic]

Alternatively, it looks like std::exp and std::log can be used, which are overloaded to have the proper behavior for each type. I've updated the PR to use those.

[Rossmaxx]

LGTM.