"How is it even possible for a single fixed measurer to determine both RSSI and Noise?" - very good question. The noise they are talking about is receiver noise and not interfering signal. At very low powers, the noise is mostly the thermal noise of the receiver: ie, if you were to disconnect the antenna and replace it with a 50 Ohm load (most RF systems are 50 Ohm) you will measure a certain level of noise. So, even if you had all the ideal components, your noise power would be P = k*T*B*G, where k is the Boltzmann's constant, T is the temperature in K, B is the bandwidth in Hz, and G is the gain of your system. In reality, every component adds noise as specified by its noise figure (listed in the datasheet of every RF component).
If you look again at the noise power equation, you will see that by reducing bandwidth, you also reduce the noise. However, high bandwidth is necessary for high data rates, which explains why you need good SNR for high data rates.
"Why both values are negative and measured in dBm" - 0 dBm means the power is 1 mW. -20 dbm means the power is .01 mW. The minus indicates the number of dB below 0 dBm. Without the minus, it would have been above 0 dBm
"But who radiate that power?" - in case of noise, it is internal, in case of signal, the transmitter. However, fundamentally it doesn't matter.
"But why is its value so small then?" - it comes from what is called Friis transmission formula. So, with several simplifications, imagine that my transmit antenna radiates power isotropically in all directions. So, your power is uniformly distributed on the surface of a sphere of radius r (and surface area 4*pi*r^2), where r is the distance from the transmit antenna. In Imagine, that your receive antenna is about 1 m^2 and it can capture all the radiation that hits its surface. Now, it can only capture 1/(4*pi*r^2) of all the radiation, making the receive power very tiny and RF engineering a complex field :). This is a very hand wavy explanation but I hope it makes sense
Spehro's comment is basically the answer. Imagine a link which subtracts a random, varying noise value between 0 and 0.5V from the input. You measure 4.5V at one end, but that could mean either a 4.5V input with no noise or 5V minus noise. You can't tell, and have lost your signal entirely.
With a digital signal varying between 5V and 0V over the same link, you can declare everything over 2.5V to be a '1', and lose no signal.
Best Answer
You're assuming there is such a difference and that it is related to the microphone and ADC.
Background noise is not so much eliminated by the microphone or ADC. To suppress background noise, multiple microphones are used and the actual suppression is done by digital signal processing, either in software or in a DSP.
Audio ADCs are now so cheap to make on a chip that it is not worth the effort to make a lower quality version. It does not save any money so no phone manufacturer skimps on the audio ADC.
With microphones it is similar, for speech almost any cheap microphone does the job for use in a phone. For a studio recording microphone its a different story of course but then size and cost are less of an issue.
If you were able to look into the designs of smartphones you would see that they're all very similar. Even though the parts are different, the specifications for the components will be very similar.
Phone manufacturers consider their designs proprietary information and they will not easily share it so that is why it is difficult to find this information.