802.11x is significantly more complicated than Zigbee, and the TCP/IP stack you need to make it work on top of all that is similarly complex. If you purchase something like the Lantronix WiPort or Digi Connect WiMe to make 802.11x networking as easy as communicating over a serial port, you're paying for a lot of stuff (there's a whole ARM based server in those modules!) that you don't need if you're willing to do a lot of RF hardware design and software integration.
If you're interested in taking that path, you should get some background understanding of the IEEE 802.11x network architecture. This is one of just a couple IEEE standards available for free through the IEEE-Get program.
Once you've got an overview of the networking system, look up the Maxim MAX283X line of transcievers. From the datasheet,
The fully integrated transceivers
include a receive path, transmit path,
voltage-controlled oscillator (VCO),
sigma-delta fractional-N synthesizer,
crystal oscillator, RSSI, PA power
detector (MAX2831), temperature
sensor, Rx and Tx I/Q error-detection
circuitry, basebandcontrol interface
and linear power amplifier (MAX2831).
The only additional components
required to implement a complete radio
front-end solution are a crystal, a
pair of baluns, a BPF, a switch, and a
small number of passive components
(RCs, no inductors required).
That's about as close as you'll get to a generic 2.4GHz transceiver that can be sanely integrated into an 802.11x network.
They cost about $5 in small quatities at the moment. As you requested, these chips only implement the PHY layer of the protocol. You still need to handle the data link layer (MAC and LLC), network layer, and transport layer on top of that before you can begin communicating at the application layer level.
Why does not these waves influence each other (they might even cancel themselves)
This absolutely happens. In fact, much of the strength of 802.11N is derived from its ability to use MIMO to take advantage of the multiple paths that the environment provides.
The latency issue is (partially) solved by using Automatic Gain Control (AGC).
Best Answer
Essentially any data radio's hardware is capable of doing this, as emitting a continuous carrier wave is one of the key parts of regulatory acceptance testing.
How you do it is not always a public feature, sometimes it requires an alternate firmware you have to obtain from the silicon manufacturer.
If you just need a low power carrier, the RF registers of the nRF24 series are very well documented and can be easily controlled from an MCU of choice with firmware you can quickly write from scratch. This may be easier than trying to get a more complex wifi-type radio to do it.