Raw conversion applications supplied by the camera manufacturers do have profiles for each camera built in, otherwise they couldn't render a remotely accurate image from the raw data at all. These profiles are created by the camera manufacturers alongside the demosaicing algorithms for each specific camera.

The same is true of large commercial raw conversion applications, such as the line of Adobe products that use Adobe Camera Raw (ACR) to convert raw data from a camera into a viewable image. There are camera profiles for each supported camera already built in. Just as with demosaicing algorithms, such profiles may be supplied by the camera maker, may be based on information supplied by the camera maker to the producers of the third party software product who create their own profile, or they may be created by the software producer based on reverse engineering of the output from the camera shooting standardized test charts with no input from the camera's manufacturer.

As the answer to the question cited in the opening paragraph of the question states, please note that when using raw data, "... the full color range supported by the camera sensor will be available, which in modern digital cameras can greatly exceed the sRGB or Adobe RGB gamut." This would also include pretty much any other gamut used for display devices or printers.

When one imports a raw image into one of these manufacturer's raw convertors or commercial raw processing software the source of the image (what kind of camera took the picture) is identified and the profile for that camera is applied. All of this happens in the background without any user interaction. If a profile for the source camera can not be found the application will usually report to the user that the file is not supported by the application. The emphasis is on letting the camera manufacturer/software supplier select their choice of the best profile for a particular camera and applying it without the need for any interaction from the end user. If such an application does allow the end user to create and/or use a different camera profile, the source of the profile and how well it works is the responsibility of the end user, not the camera maker or software supplier.

When it comes to applications such as UFRaw or Rawtherapee the situation is a bit different. These "open source/GNU General Public License" types of applications usually allow more under the hood tinkering by the end user. As such, they have the ability to attempt to render image data even when the source of that data is unknown. They allow the end user to supply, or even create, a camera profile for an image file when the source camera can not be identified by the application, when the source is identified as one for which no profile exists in the application's library of camera profiles, or even when the source camera is properly identified and a profile is available but the user wishes to apply a different profile instead.

This degree of control also allows the option of applying one of several available profiles to images from a particular source. As with many things, giving the end user more choices can also allow the end user to make mistakes that the end user will then blame on the entity who gave them the ability to make a bad choice or even blame the source of the image data to which they applied an improper color profile.

Camera manufacturers should definitely have all the information necessary to create camera color profiles for their models but it seems to me they don't provide camera profiles for their cameras. What is the reason for this?

They do provide camera profiles for their cameras inside their own image processing products. Some of them also provide such profiles to other third party software publishers. They just don't usually make them publically available. For why this is the case you would need to ask the people who make such decisions for each specific camera manufacturer. One can guess that trade secrets and protecting proprietary information¹ might be one motivation. Another might be to prevent users from inappropriately using them and then blaming the manufacturer for creating a bad product. But the question can only be definitively answered by those who make such decisions.

¹Just because something can be "cracked" by someone determined enough and with enough resources doesn't mean the owners of proprietary information won't make it as difficult as possible to do so.