The short answer is, at the moment, only for the Moon. I include below a brief description of the JPL DE format to arrive to this conclusion.
This link explains quite well the format of JPL DE ephemerides, which are available here. The link is for the DE440, but a directory up we can find previous ones also.
In summary, we need 2 files to properly understand them: the ephemerides data themselves (split into multiple files, the ascpXXXX.440 in the previous link; the XXXX denotes the starting year for the data in that file) and the header (header.440). The header contains quite a lot of information, but the most interesting part for this discussion are the 3 lines that come after the line "GROUP 1050".
These 3 lines contain information about the 15 types of records that are currently included in the DE (these 15 types of records are given in Table 1 here. The 2nd line specifies the number of Chebyshev coefficients given for each property of each of the records (e.g., for position records, we have 3 properties, X, Y and Z, and each of these has the number of Chebyshev coefficients given in this line). The 3rd line specifies the so-called "subintervals" with which each property is defined, which determines the time resolution (each complete data block of the ephemerides is an interval of 32 days, so if the number of subintervals is 1, we just get the coefficients defined for that period of 32 days; if it is 2, we get values for them for a period of 16 days and then different values for the next period of 16 days, and so on).
As can be seen from the linked Table 1, records 1 to 11 are positions. Record 12 are nutations of Earth (2 components), record 13 are lunar librations (3 components, the 3 angles), record 14 are lunar mantle angular velocities (also 3 components), and record 15 is the difference in seconds between TT and TDB.
So if we go to line 2 after the mentioned "GROUP 1050" line in the header, we can know how many coefficients are given for each of these properties. Properties for which values are not given have a value of 0 here. For DE440, we have 10 coefficients for Earth nutations (with 4 subintervals), and also 10 coefficients for each of the components of lunar libration (with 4 subintervals also). So we have indeed values for lunar librations! But it looks like that's about it in terms of librations, since librations for other celestial bodies are not even listed (might be worth saying, yet) in the list of defined records.
Lunar mantle angular velocity has 0 coefficients (i.e., no values given), for the time being!
As a side note, as you noted in comments, I thought it is interesting that more information is given about the Moon than about the Earth. A summary for each in DE440:
- For Moon: positions with time resolution of 4 days and 13 coefficients for each component, librations with time resolution of 8 days and 10 coefficients for each component
- For Earth: positions with time resolution of 16 days and 14 coefficients for each component (and in fact, they are positions of the Earth-Moon barycenter! so Earth position must be derived considering the Earth-Moon mass ratio, which is also given in the header file by the way), nutations with time resolution of 8 days and 10 coefficients for each component (but we only have 2 components here compared to the 3 components for Moon librations, and there seems to be provisions for even more Moon information, i.e., mantle angular velocity)