You are correct in your understanding "For example, Routing protocols such as OSPF, IS-IS, and BGP are handled by RE. Inline BFD is handled by PFE. Right?" [though Inline BFD handling is platform and junos version specific.
All Juniper Networks devices that run on the Junos operating system share the same common design philosophy, which is to have a clean separation of the control and forwarding planes. In the high-end devices (for example, M-series routers, MX edge devices, and Data Center SRXs), this separation is created in hardware, whereas the other devices (J-series routers and Branch Office SRXs) maintain this division in software. The forwarding plane is referred to as the Packet Forwarding Engine (PFE), and the control plane is called the Routing Engine (RE).
The RE’s primary functions are to manage the PFE, control the device’s software (Junos operating system), manage the command-line interface (CLI), provide troubleshooting tools, and maintain the route tables (both the route table and the route forwarding table). The forwarding table, a subset of the route table, is passed down to the PFE and is used to forward traffic. In this way, the RE never has to be directly involved in packet forwarding, which allows more resources for the actual control functions.
You may also visit the below mentioned link for a detailed explanation:
Table 1(Default Queue Assignments for Packets Generated by the Routing Engine) shows that BFD, LACP, VRRP, and RPM are handled by the Routing Engine, which is sure to be PFE processing is possible via distributed ppmmd?
The link you shared in the above post explains how traffic generated by the routing engine are assigned to different queues available.
By default, PPM will ensure it handles the packet processing for it's client processes at the linecard level. Only if the linecard CPU is busy, Routing engine will come into the picture and handle packet processing.
"The responsibility for PPM processing on the switch is distributed between the Routing Engine and the access interfaces for all protocols that use PPM by default. This distributed model provides a faster response time for protocols that use PPM than the response time provided by the nondistributed model."