Also why they say that we can have one learning domain in a bridge domain versus many learning domains in a bridge domain?
Bridge Domains contain Learning Domains, they are not mutually exclusive.
What does this mean and whats the use case?
First, let's look at a Bridge Domain with a single Learning Domain. This effectively means that you have a single broadcast domain, similar to a single VLAN. Even if you had multiple interfaces with multiple VLANs, they would still learn the same way. In short, learning is done purely on the 6-octet MAC address, regardless of any L2 specifics (i.e. VLAN tags). This is also known as "unqualified learning". For example, if you had MAC AA:BB:CC:DD:EE:FF on both VLANs 100 and 500, only one entry would be learned.
Now let's look at a Bridge Domain with multiple Learning Domains. Instead of learning purely on the MAC address, the multiple learning domains split any VLANs into their own learning domain. In short, learning is done based on both VLAN tag(s) and MAC address. Using the same example, if you had AA:BB:CC:DD:EE:FF on both VLANs 100 and 500, you'd have two entries, one for each VLAN.
The book's statement of "The primary use case for multiple learning domains is for Service Providers to tunnel customer VLANs" is accurate, just not as specific as it could be. It's referring to VPLS services, in the case of a single learning domain it would learn as a single VPLS instance, whereas multiple learning domains, you would have each VLAN map to its own VPLS.
RFC4761 - Section 4.2.6 gives us a good reference for this behavior:
The key for normal Ethernet MAC learning is usually just the
(6-octet) MAC address. This is called "unqualified learning".
However, it is also possible that the key for learning includes the
VLAN tag when present; this is called "qualified learning".